Agent-Skills

Create, review, and validate an alignment checkpoint. Use when a request is ambiguous, high-stakes, multi-step, or requires explicit approval before tool use.

Runs approved Harness Engineering plans in recurring phases: check live state, confirm continuation authority, execute only the active he-work slice, verify gates, update Linear only with approval, and stop before unsafe staging or closure. Use when a plan, issue, or PR needs continued phase-by-phase execution.

Scaffold Codex plugin packages with deterministic manifests, marketplace metadata, and traceability or evidence contracts for non-trivial adoption. Use when creating plugin roots or adopting existing skills into plugin ownership.

Run a bounded Harness Engineering lifecycle across multiple stages. Use when the user wants coordinated brainstorm, spec, plan, work, review, and fix flow rather than one isolated stage.

Use when hardening, converting, auditing, or pre-release checking a Codex plugin package by verifying manifest paths, bundled skills, hooks, MCP/app config, validation gates, and release blockers.

Analyze and validate compound Harness Engineering run state, blockers, validation status, and Linear context. Use when lifecycle runs drift, gates fail, blockers appear, or compound work needs refresh.

Routes old phase-heartbeat requests into HE phase work: check the approved plan, reuse or create a 10-minute continuation only with authority, run phase gates, delegate the active slice to he-work, and stop before commit or tracker closure. Use when a user says phase heartbeat, keep this PR moving, or continue phases.

Analyze, fix, and validate Biome linting workflows. Use when JavaScript or TypeScript projects need Biome commands, diagnostics, safe fixes, or CI lint gates.

Analyze repo, tracker, PR, validation, session, and .harness evidence. Use when multi-stage Harness Engineering work needs safe resume routing.

Install validated Codex plugins from trusted sources with quarantine validation, provenance, and rollback. Use when distribution and installation are the primary goals.

Use when reviewing, creating, shrinking, or refactoring AGENTS.md and directly linked instruction guidance that need scoped routing, deduplication, contradiction resolution, or progressive disclosure.

Analyzes Codex skill-management requests, selects the workflow lane, and returns selected_lane, mode, next_step, and blockers. Use when the user says create a skill, add/update/fix/review a skill, install/sync/list skills, choose a workflow, or merge/retire a skill.

Restore broken behavior by reproducing failures, identifying root cause, and delivering verified fixes. Use when the user needs regression debugging, incident triage, or bug repair from tracker or direct reports.

Run bounded automated experiment iterations by recording baselines, applying hypothesis patches, comparing metrics, protecting regression guards, and deciding keep, discard, rollback, or block. Use when automated research is requested or a repo/skill needs evidence-backed research, metric tracking, or safe optimisation loops.

Generate closure-grade HE eval and drift proof for one execution slice. Use when Linear, milestone, or source-prompt closure needs validation evidence.

Deepen an existing implementation plan so sequencing, verification, and risk treatment are strong enough for execution. Use when the user wants Harness Engineering plan hardening before he-work.

Plan execution work from specs, brainstorm outputs, bugs, or feature requests into an implementation-ready sequence. Use when the user needs the Harness Engineering planning stage before execution.

Capture a completed Codex workflow as a reusable SKILL.md package by analyzing session context plus optional session-collector evidence, interviewing the user with structured prompts, and writing a validated skill artifact. Use when the user asks to skillify or operationalize a repeatable process.

Define problem scope, requirements, and decision options before spec or plan stages. Use when the user has ambiguity in what to build, why it matters, or which direction to choose.

Explore Harness Engineering options, filter ornate or weak ideas, recover dropped leverage, and select survivor routes before commitment. Use when intent, stage choice, tradeoffs, idea quality, or possible solution shapes are still unsettled before spec, plan, Linear, or implementation work.

Reviews and improves SKILL.md packages by fixing audit findings, triggers, examples, evals, token budget, release proof, safety verdicts, comparator/baseline choices, and bounded code-lens hardening. Use when the user says improve a skill, fix a skill file, review SKILL.md, raise Tessl score, reduce context cost, add skill evals, or prepare a plugin skill for release.

Create, diagnose, and validate a local dev bootstrap. Use when the user asks to clone a repo, install toolchains, install dependencies, and prove the project runs.

Analyze broad, mixed, or unclear Plugin Factory follow-up requests and select the correct plugin lane. Use when plugin intent lacks a clear lane owner.

Scan Codex session history for skill failures, usage patterns, and coverage gaps. Use when the user wants daily skill-health monitoring or evidence-backed recommendations about installing, improving, merging, or pruning skills.

Apply approved fixes for unresolved CodeRabbit review comments, Codex P1-P3 findings, PR feedback, and code review issues with validation evidence. Use when asked to address review comments, fix review findings, clear unresolved comments, or autofix PR feedback.