Skill Claude CodeCodex
Use when reviewing, validating, or cleaning up an AGENTS.md file, especially for contradictions, unclear instructions, stale harness-specific tool names, or unsafe automation guidance.
Skill Claude CodeCodex
Use when the user asks to commit, create a git commit, stage and commit changes, or wants a gitmoji commit message for staged changes.
Skill Claude CodeCodex
Use when you need to audit an npm package, a GitHub repository, or pending dependency updates for supply-chain malware and risky lifecycle scripts before installation. Default action without parameters is to audit all pending updates.
Skill Claude CodeCodex
Use when migrating, splitting, or restructuring agent-specific instructions from CLAUDE.md into AGENTS.md, especially when preserving project documentation while moving agent guidance.
Skill Claude CodeCodex
Use when the user asks for GitHub PR review findings, review comments, unresolved findings, bot feedback, or reviewer feedback grouped by severity.
Skill Claude CodeCodex
Audit, clean, and simulate the rule-matching of pi-bash-approval whitelist rules. Use when a command is unexpectedly blocked, when you want to verify if a new regex rule is secure, or to clean up duplicates and redundant rules from /.pi/agent/.bash-approval.
Skill Claude CodeCodex
Use when testing, inspecting, or interacting with web pages through the browser tools extension in the pi-plugins package.
Skill Claude CodeCodex
Use when you need to stress-test a plan against the project's existing domain model (CONTEXT.md) and architectural decisions (ADRs). Challenges terminology, aligns with canonical language, and updates documentation inline.
Skill Claude CodeCodex
Use when you need to compact the current conversation into a document for another agent or session. Summarizes context, decisions, and next steps.
Skill Claude CodeCodex
Use when you want to surface architectural friction, find refactoring opportunities, consolidate tightly-coupled modules, or propose deepening opportunities (turning shallow modules into deep ones). Informed by CONTEXT.md and ADRs. Aimed at testability and AI-navigability.