From 7865f76ce657d1b538cbae7b2c2307a1735da988 Mon Sep 17 00:00:00 2001 From: Trevin Chow Date: Sat, 18 Apr 2026 16:52:43 -0700 Subject: [PATCH 1/2] fix(ce-work): reject plan re-scoping into human-time phases The skill had no explicit guardrail against the agent estimating human-hours per implementation unit and proposing multi-day session breakdowns. Tightens Phase 1 Step 1 so "approval" only applies when clarifications were needed, and adds a Common Pitfall that points at subagent dispatch (Phase 1 Step 4) as the correct lever for large plans. Synced to ce-work-beta. --- plugins/compound-engineering/skills/ce-work-beta/SKILL.md | 3 ++- plugins/compound-engineering/skills/ce-work/SKILL.md | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/plugins/compound-engineering/skills/ce-work-beta/SKILL.md b/plugins/compound-engineering/skills/ce-work-beta/SKILL.md index 0eb0bc14b..175c8218b 100644 --- a/plugins/compound-engineering/skills/ce-work-beta/SKILL.md +++ b/plugins/compound-engineering/skills/ce-work-beta/SKILL.md @@ -108,7 +108,7 @@ Determine how to proceed based on what was provided in ``. - Review any references or links provided in the plan - If the user explicitly asks for TDD, test-first, or characterization-first execution in this session, honor that request even if the plan has no `Execution note` - If anything is unclear or ambiguous, ask clarifying questions now - - Get user approval to proceed + - If clarifying questions were needed above, get user approval on the resolved answers. If no clarifications were needed, proceed without a separate approval step — plan scope is the plan's authority, not something to renegotiate - **Do not skip this** - better to ask questions now than build the wrong thing 2. **Setup Environment** @@ -414,3 +414,4 @@ When `delegation_active` is true after argument parsing, read `references/codex- - **Forgetting to track progress** - Update task status as you go or lose track of what's done - **80% done syndrome** - Finish the feature, don't move on early - **Skipping review** - Every change gets reviewed; only the depth varies +- **Re-scoping the plan into human-time phases** - The plan's Implementation Units define the scope of execution. Do not estimate human-hours per unit, propose multi-day breakdowns, or ask the user to pick a subset of units for "this session". Agents execute at agent speed, and context-window pressure is addressed by subagent dispatch (Phase 1 Step 4), not by phased sessions. If the plan truly is too large, Phase 0's Large routing already sends it back to `/ce-brainstorm` or `/ce-plan` diff --git a/plugins/compound-engineering/skills/ce-work/SKILL.md b/plugins/compound-engineering/skills/ce-work/SKILL.md index f76113a19..6aaa42f42 100644 --- a/plugins/compound-engineering/skills/ce-work/SKILL.md +++ b/plugins/compound-engineering/skills/ce-work/SKILL.md @@ -55,7 +55,7 @@ Determine how to proceed based on what was provided in ``. - Review any references or links provided in the plan - If the user explicitly asks for TDD, test-first, or characterization-first execution in this session, honor that request even if the plan has no `Execution note` - If anything is unclear or ambiguous, ask clarifying questions now - - Get user approval to proceed + - If clarifying questions were needed above, get user approval on the resolved answers. If no clarifications were needed, proceed without a separate approval step — plan scope is the plan's authority, not something to renegotiate - **Do not skip this** - better to ask questions now than build the wrong thing 2. **Setup Environment** @@ -341,3 +341,4 @@ When all Phase 2 tasks are complete and execution transitions to quality check, - **Forgetting to track progress** - Update task status as you go or lose track of what's done - **80% done syndrome** - Finish the feature, don't move on early - **Skipping review** - Every change gets reviewed; only the depth varies +- **Re-scoping the plan into human-time phases** - The plan's Implementation Units define the scope of execution. Do not estimate human-hours per unit, propose multi-day breakdowns, or ask the user to pick a subset of units for "this session". Agents execute at agent speed, and context-window pressure is addressed by subagent dispatch (Phase 1 Step 4), not by phased sessions. If the plan truly is too large, Phase 0's Large routing already sends it back to `/ce-brainstorm` or `/ce-plan` From 97a2fbfd918ff6b0748060bf656c8ee0e61dc971 Mon Sep 17 00:00:00 2001 From: Trevin Chow Date: Sat, 18 Apr 2026 17:12:38 -0700 Subject: [PATCH 2/2] Address PR review feedback (#600) - Narrow the oversized-plan escape hatch reference: Phase 0's Large routing only fires for bare-prompt input, since plan-file input explicitly skips Phase 0. Make the rule explicit for plan-file input (escalate to /ce-plan) and preserve the bare-prompt routing reference for that path. --- plugins/compound-engineering/skills/ce-work-beta/SKILL.md | 2 +- plugins/compound-engineering/skills/ce-work/SKILL.md | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/plugins/compound-engineering/skills/ce-work-beta/SKILL.md b/plugins/compound-engineering/skills/ce-work-beta/SKILL.md index 175c8218b..486605f9c 100644 --- a/plugins/compound-engineering/skills/ce-work-beta/SKILL.md +++ b/plugins/compound-engineering/skills/ce-work-beta/SKILL.md @@ -414,4 +414,4 @@ When `delegation_active` is true after argument parsing, read `references/codex- - **Forgetting to track progress** - Update task status as you go or lose track of what's done - **80% done syndrome** - Finish the feature, don't move on early - **Skipping review** - Every change gets reviewed; only the depth varies -- **Re-scoping the plan into human-time phases** - The plan's Implementation Units define the scope of execution. Do not estimate human-hours per unit, propose multi-day breakdowns, or ask the user to pick a subset of units for "this session". Agents execute at agent speed, and context-window pressure is addressed by subagent dispatch (Phase 1 Step 4), not by phased sessions. If the plan truly is too large, Phase 0's Large routing already sends it back to `/ce-brainstorm` or `/ce-plan` +- **Re-scoping the plan into human-time phases** - The plan's Implementation Units define the scope of execution. Do not estimate human-hours per unit, propose multi-day breakdowns, or ask the user to pick a subset of units for "this session". Agents execute at agent speed, and context-window pressure is addressed by subagent dispatch (Phase 1 Step 4), not by phased sessions. If a plan-file input is genuinely too large for a single execution, say so plainly and suggest the user return to `/ce-plan` to reduce scope — don't invent session phases as a workaround. For bare-prompt input, Phase 0's Large routing already handles oversized work diff --git a/plugins/compound-engineering/skills/ce-work/SKILL.md b/plugins/compound-engineering/skills/ce-work/SKILL.md index 6aaa42f42..f36ddef43 100644 --- a/plugins/compound-engineering/skills/ce-work/SKILL.md +++ b/plugins/compound-engineering/skills/ce-work/SKILL.md @@ -341,4 +341,4 @@ When all Phase 2 tasks are complete and execution transitions to quality check, - **Forgetting to track progress** - Update task status as you go or lose track of what's done - **80% done syndrome** - Finish the feature, don't move on early - **Skipping review** - Every change gets reviewed; only the depth varies -- **Re-scoping the plan into human-time phases** - The plan's Implementation Units define the scope of execution. Do not estimate human-hours per unit, propose multi-day breakdowns, or ask the user to pick a subset of units for "this session". Agents execute at agent speed, and context-window pressure is addressed by subagent dispatch (Phase 1 Step 4), not by phased sessions. If the plan truly is too large, Phase 0's Large routing already sends it back to `/ce-brainstorm` or `/ce-plan` +- **Re-scoping the plan into human-time phases** - The plan's Implementation Units define the scope of execution. Do not estimate human-hours per unit, propose multi-day breakdowns, or ask the user to pick a subset of units for "this session". Agents execute at agent speed, and context-window pressure is addressed by subagent dispatch (Phase 1 Step 4), not by phased sessions. If a plan-file input is genuinely too large for a single execution, say so plainly and suggest the user return to `/ce-plan` to reduce scope — don't invent session phases as a workaround. For bare-prompt input, Phase 0's Large routing already handles oversized work