Enforce level-aware skill capability contracts and agent-safe projections#4345
Conversation
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit PR Review — Verdict: approved
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit review passed. No blocking issues found.
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit PR Review — Verdict: changes_requested
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit review: blocking findings detected. See inline comments.
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit PR Review — Verdict: changes_requested
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit review: blocking findings detected. See inline comments — changes required.
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit PR Review — Verdict: changes_requested
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit review — Verdict: changes_requested. Blocking and warning findings were detected; see the inline comments for required changes.
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit PR Review — Verdict: changes_requested
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit review: blocking findings detected. See inline comments — automated verdict: changes_requested. GitHub does not allow an author to request changes on their own PR.
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit PR Review — Verdict: approved_with_comments
Trecek
left a comment
There was a problem hiding this comment.
AutoSkillit review: warning-only findings detected. See inline comments — no blocking changes required.
0799ffc to
b2a106f
Compare
Summary
This change separates machine-authoritative skill capability contracts from model-facing skill documents, validates exact execution-role ownership, resolves project-local overrides consistently, and persists the effective contract and projected document across resumed sessions. It also closes the follow-up audit findings by using effective project-root resolution during LLM triage and doctor checks, aligning projection and subprocess
cwd, and adding regression coverage for override identity, role validation, fresh-versus-resume guards, and contract persistence.Requirements
Closes #4332
Implementation Plan
Plan file:
/home/talon/projects/generic_automation_mcp/.autoskillit/temp/rectify/rectify_run_skill_capability_contract_immunity_2026-07-22_201946.md🤖 Generated with Claude Code via AutoSkillit