0.32.0

Bug Fixes 🐛

• (action) Install Claude Code only for Claude runtime by @tmustier in #334
• (output) Render finding evidence traces by @dcramer in #335

Internal Changes 🔧

Evals

• Upgrade vitest-evals to 0.9.0-beta.6 by @sentry-junior in #332
• Move eval harness into workspace package by @dcramer in #331
• Add vitest-evals reporter bug fixtures by @dcramer in #330
• Centralize fixture provenance by @dcramer in #329

Other

• (release) Create source tags before artifact tags by @dcramer in #328

0.31.1

Bug Fixes 🐛

• (action) Install Claude Code only for Claude runtime by @tmustier in #334
• (output) Render finding evidence traces by @dcramer in #335

Internal Changes 🔧

Evals

• Upgrade vitest-evals to 0.9.0-beta.6 by @sentry-junior in #332
• Move eval harness into workspace package by @dcramer in #331
• Add vitest-evals reporter bug fixtures by @dcramer in #330
• Centralize fixture provenance by @dcramer in #329

Other

• (release) Create source tags before artifact tags by @dcramer in #328

0.31.0

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Add @file target lists by @dcramer in #316
• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add Pi runtime by @dcramer in #318
• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Honor PR trigger parallelism by @dcramer in #325
• Preload Pi node-only providers by @dcramer in #324
• Validate Pi model selectors before execution by @dcramer in #321
• Correct fix evaluation verdict handling by @dcramer in #314
• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Fill finding analytics gaps by @dcramer in #311
• Align attributes with OTel semantics by @dcramer in #307
• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (code-review) Calibrate config severity by @dcramer in [#319](https://github.com/getsentry/warden/pull/...

0.30.0

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Add @file target lists by @dcramer in #316
• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add Pi runtime by @dcramer in #318
• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Honor PR trigger parallelism by @dcramer in #325
• Preload Pi node-only providers by @dcramer in #324
• Validate Pi model selectors before execution by @dcramer in #321
• Correct fix evaluation verdict handling by @dcramer in #314
• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Fill finding analytics gaps by @dcramer in #311
• Align attributes with OTel semantics by @dcramer in #307
• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (code-review) Calibrate config severity by @dcramer in [#319](https://github.com/getsentry/warden/pull/...

0.29.1

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Add @file target lists by @dcramer in #316
• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add Pi runtime by @dcramer in #318
• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Validate Pi model selectors before execution by @dcramer in #321
• Correct fix evaluation verdict handling by @dcramer in #314
• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Fill finding analytics gaps by @dcramer in #311
• Align attributes with OTel semantics by @dcramer in #307
• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (code-review) Calibrate config severity by @dcramer in #319
• (evals) Update default model to claude-sonnet-4-6 by @dcramer in #163
• (o11y) Add gen_ai.request.model attribute by @obostjancic in [#247](https...

0.29.0

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Add @file target lists by @dcramer in #316
• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add Pi runtime by @dcramer in #318
• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Correct fix evaluation verdict handling by @dcramer in #314
• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Fill finding analytics gaps by @dcramer in #311
• Align attributes with OTel semantics by @dcramer in #307
• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (code-review) Calibrate config severity by @dcramer in #319
• (evals) Update default model to claude-sonnet-4-6 by @dcramer in #163
• (o11y) Add gen_ai.request.model attribute by @obostjancic in #247
• (prepare) Skip files with zero content automatically by @dcramer in [#22...

0.28.0

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Correct fix evaluation verdict handling by @dcramer in #314
• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Fill finding analytics gaps by @dcramer in #311
• Align attributes with OTel semantics by @dcramer in #307
• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (evals) Update default model to claude-sonnet-4-6 by @dcramer in #163
• (o11y) Add gen_ai.request.model attribute by @obostjancic in #247
• (prepare) Skip files with zero content automatically by @dcramer in #226
• (review) Preserve active Warden threads by @dcramer in #272
• (reviews) Compact inline finding comments by @dcramer in #315
• (sentry) Tag ac...

0.27.0

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Fill finding analytics gaps by @dcramer in #311
• Align attributes with OTel semantics by @dcramer in #307
• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (evals) Update default model to claude-sonnet-4-6 by @dcramer in #163
• (o11y) Add gen_ai.request.model attribute by @obostjancic in #247
• (prepare) Skip files with zero content automatically by @dcramer in #226
• (review) Preserve active Warden threads by @dcramer in #272
• (sentry) Tag action errors with repository by @dcramer in #303
• (skill-builder) Stabilize generated skill authoring by @dcramer in #284
• (s...

0.26.2

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• Render GitHub footers as muted text by @dcramer in #306
• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (evals) Update default model to claude-sonnet-4-6 by @dcramer in #163
• (o11y) Add gen_ai.request.model attribute by @obostjancic in #247
• (prepare) Skip files with zero content automatically by @dcramer in #226
• (review) Preserve active Warden threads by @dcramer in #272
• (sentry) Tag action errors with repository by @dcramer in #303
• (skill-builder) Stabilize generated skill authoring by @dcramer in #284
• (skills) Support private repos in warden add --remote by @dcramer in #187
• (usage) Normalize inputTokens to include cached and cache-write tokens by @vgrozd...

0.26.1

New Features ✨

Action

• Support org base configs with repo overlays by @dcramer in #279
• Write structured findings JSON and expose as output by @gricha in #233

Cli

• Build generated skills from path targets by @dcramer in #282
• Add --staged flag and default to HEAD for uncommitted changes by @dcramer in #205
• Display confidence level in findings output by @dcramer in #194
• Always-on JSONL logging with repo-local log directory by @dcramer in #179
• Add --fail-fast flag to stop after first finding by @dcramer in #174
• Surface failure details for chunk analysis and finding extraction by @dcramer in #162
• Interactive fix step-through with y/n/a/s/q prompt by @dcramer in #161

Docs

• Improved landing visuals by @PickleNik in #245
• Add LLM content negotiation and llms.txt by @dcramer in #178

Runs

• Write chunk-level JSONL logs by @dcramer in #265
• Stream in-progress runs via 'warden runs follow' by @dcramer in #261

Sdk

• Add finding verification pass by @dcramer in #290
• Store Claude SDK sessions in .warden/sessions/ by @dcramer in #193

Skills

• Add built-in code review skill by @dcramer in #298
• Add built-in security review skill by @dcramer in #292
• Consolidate warden-sweep into bundled scripts with prescribed reporting by @dcramer in #201
• Replace notseer with find-warden-bugs skill by @dcramer in #200
• Add warden-sweep skill and bundled skill installation by @dcramer in #196
• Add agent discovery alongside skills by @dcramer in #136

Sweep

• Add --skill flag to scan.py by @dcramer in #211
• Add master tracking issue for sweep PRs by @dcramer in #210

Telemetry

• Add skill and model attribution to metrics by @gricha in #231
• Add namespaced fix-eval feedback metrics by @dcramer in #229
• Add global scope attributes, run metric, and trace ID surfacing by @dcramer in #166
• Add input/output attributes to gen_ai spans by @dcramer in #142
• Add per-turn tracing for Claude Code SDK agent calls by @dcramer in #141

Other

• (config) Make auxiliary max retries configurable via defaults.auxiliaryMaxRetries by @dcramer in #175
• (dedup) Add intra-batch finding consolidation by @dcramer in #135
• (init) Add section-based output and interactive skill prompt by @dcramer in #182
• (o11y) Capture run failures in JSONL output by @dcramer in #259
• (output) Default to COMMENT instead of REQUEST_CHANGES on PRs by @dcramer in #137
• (runner) Global concurrency pool for file analyses by @dcramer in #158
• (skill-builder) Build generated repo-local skills by @dcramer in #280
• Add verification field and confidence filtering by @dcramer in #191
• Pnpm workspace monorepo with docs migration by @dcramer in #185
• Adopt dotagents for skill management by @dcramer in #168
• Cross-location finding merge by @dcramer in #148
• Add env vars to generated workflow and docs by @dcramer in #138
• Add Sentry observability (errors, tracing, metrics, structured logs) by @dcramer in #133
• Split failOn into independent requestChanges and failCheck controls by @dcramer in #129

Bug Fixes 🐛

Action

• Use relative findings output path by @dcramer in #289
• Handle missing warden.toml gracefully by @gricha in #235
• Thread maxContextFiles config through Action workflows by @dcramer in #216

Cli

• Label built-in skill sources by @dcramer in #294
• Report final per-file finding counts by @dcramer in #293
• Align local review diff context by @dcramer in #276
• Replace process.exit(130) with thrown error for graceful cleanup by @dcramer in #213
• Write --output file on auth error early exit by @dcramer in #214
• Normalize path separators in synthetic file changes by @dcramer in #217
• Exclude node_modules from gitignore file discovery by @dcramer in #220
• Collapse skipped files into a single count on Ctrl+C by @dcramer in #172
• Render warnings through Ink's Static component by @dcramer in #170
• Cap 'Other Files' list in prompts for CLI mode by @dcramer in #171
• Move completed files to Static to prevent flickering by @dcramer in #149
• Run all skills locally regardless of trigger type by @gricha in #145
• Resolve remote config for --skill flag by @dcramer in #130

Config

• Allow base configs to use built-in skills by @dcramer in #305
• Warn on duplicate layered skills by @dcramer in #304
• Add missing fields to TOML writer for config round-trip by @dcramer in #218

Output

• List only successfully applied fixes in PR body by @dcramer in #219
• Use backticks instead of brackets in attribution footer by @dcramer in #197

Sdk

• Short-circuit repeated provider failures by @dcramer in #291
• Strip invalid suggested fixes from reports by @dcramer in #228
• Prevent Warden sessions from polluting Claude Code history by @dcramer in #202
• Handle EPIPE errors from Claude subprocess failures by @dcramer in #198
• Add verification checklist and Glob tool to reduce analysis hallucinations by @dcramer in #173
• Constrain skill findings to diff hunk line range by @dcramer in #151

Sweep

• Add self-validation to fix subagent prompt by @dcramer in #223
• Branch worktrees from default branch instead of HEAD by @dcramer in #221

Telemetry

• Improve OTel GenAI semantic convention compliance by @dcramer in #140
• Fix missing Sentry traces and negative token costs by @dcramer in #139

Other

• (ci) Enable auto-generated changelogs for GitHub releases by @gricha in #234
• (evals) Update default model to claude-sonnet-4-6 by @dcramer in #163
• (o11y) Add gen_ai.request.model attribute by @obostjancic in #247
• (prepare) Skip files with zero content automatically by @dcramer in #226
• (review) Preserve active Warden threads by @dcramer in #272
• (sentry) Tag action errors with repository by @dcramer in #303
• (skill-builder) Stabilize generated skill authoring by @dcramer in #284
• (skills) Support private repos in warden add --remote by @dcramer in #187
• (usage) Normalize inputTokens to include cached and cache-write tokens by @vgrozdanic in #143
• Preserve interrupted findings and parse run...