Skip to content

Add deterministic agent tool contracts - #2020

Merged
Aaronontheweb merged 1 commit into
devfrom
goal/pit-of-success-agent-tools
Aug 20, 2026
Merged

Add deterministic agent tool contracts#2020
Aaronontheweb merged 1 commit into
devfrom
goal/pit-of-success-agent-tools

Conversation

@Aaronontheweb

@Aaronontheweb Aaronontheweb commented Aug 19, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Add the validated OpenSpec plan for safer agent tool use.
  • Add a PII-free tool-friction fixture with strict mutation and privacy tests.
  • Record the actual parent and subagent schema footprint without schema bodies.

Scope

This PR defines deterministic contracts only. Product implementation follows in the stacked PRs.

Deterministic baseline evidence

  • A fixed 200-tool synthetic catalog produced 3 parent tool definitions totaling 1,047 serialized UTF-8 bytes.
  • The same catalog produced 202 child tool definitions totaling 133,816 serialized UTF-8 bytes.
  • The evidence stores aggregate counts and byte sizes only; it contains no schema bodies, raw prompts, logs, paths, session identifiers, or production-derived tool names.
  • No hosted behavioral run was used as evidence for this PR.

Validation

  • Tool-friction contract tests: 35 passed.
  • Model-visible schema-footprint test: passed.
  • Strict OpenSpec validation: passed.
  • Header verification: passed.
  • Slopwatch: zero issues.
  • Diff check: passed.

Comment on lines +921 to +924
var path = Path.Combine(
AppContext.BaseDirectory,
"ToolFootprintEvidence",
"tool-schema-footprint-baseline.json");
=> Convert.ToHexString(SHA256.HashData(bytes)).ToLowerInvariant();

private static string FixturePath()
=> Path.Combine(AppContext.BaseDirectory, "ToolFrictionEvidence", FixtureFile);
@Aaronontheweb
Aaronontheweb enabled auto-merge (squash) August 19, 2026 20:29
@Aaronontheweb
Aaronontheweb disabled the stack merge August 19, 2026 20:35
@Aaronontheweb
Aaronontheweb force-pushed the goal/pit-of-success-agent-tools branch from d3d53b4 to 12f8c39 Compare August 19, 2026 20:44
@Aaronontheweb Aaronontheweb changed the title Add agent tool affordance baseline Add deterministic agent tool contracts Aug 19, 2026
@Aaronontheweb
Aaronontheweb force-pushed the goal/pit-of-success-agent-tools branch from 12f8c39 to b51cdce Compare August 19, 2026 21:10
@Aaronontheweb
Aaronontheweb enabled auto-merge (squash) August 19, 2026 21:19
@Aaronontheweb
Aaronontheweb force-pushed the goal/pit-of-success-agent-tools branch from b51cdce to 33daa81 Compare August 19, 2026 21:50
@Aaronontheweb
Aaronontheweb disabled the stack merge August 19, 2026 22:12
@Aaronontheweb
Aaronontheweb force-pushed the goal/pit-of-success-agent-tools branch 2 times, most recently from 4d87c6f to fd19e9c Compare August 20, 2026 04:50
@Aaronontheweb
Aaronontheweb force-pushed the goal/pit-of-success-agent-tools branch from fd19e9c to 43f9e1a Compare August 20, 2026 05:33
@Aaronontheweb
Aaronontheweb merged commit f174421 into dev Aug 20, 2026
23 checks passed
@Aaronontheweb
Aaronontheweb deleted the goal/pit-of-success-agent-tools branch August 20, 2026 08:29
@Aaronontheweb Aaronontheweb added tools Issues related to agent tools: file_read, web_search, shell_execute, image processing, etc. tests All issues related to testing, quality assurance, and smoke testing. documentation Improvements or additions to documentation evals LLM efficacy evals for Netclaw labels Aug 20, 2026
@Aaronontheweb Aaronontheweb mentioned this pull request Aug 26, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation evals LLM efficacy evals for Netclaw tests All issues related to testing, quality assurance, and smoke testing. tools Issues related to agent tools: file_read, web_search, shell_execute, image processing, etc.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant