@boshu2

pr-prep validates isolation and test coverage before generating structured pr bodies with mandatory user review, targeting external repository contributions.

release handles the full pipeline from code readiness through tagged commit and ci verification, including changelog generation, version bumps, and dry-run validation before publishing.

brainstorm clarifies goals before planning by separating intent from implementation, generating multiple approaches with tradeoff analysis, and producing bdd-shaped acceptance criteria (given/when/then) downstream for discovery and planning.

pr-implement executes scoped oss contributions via fork isolation, enforcing single-concern commits through blocking pre and post-checks against scope creep and mixed concerns.

trace reconstructs the evolution of architectural decisions by searching cass sessions, git history, handoff docs, and research artifacts, then produces a cited chronological timeline showing what changed, why, and by whom.

oss-docs audits and scaffolds documentation for open source projects, detecting language and project type to generate tier-based standard files (license, readme, contributing, coc, security, changelog). covers audit, scaffold, update, and validate workflows with troubleshooting guidance.

pr-retro extracts actionable lessons from pull request outcomes by analyzing reviewer feedback, identifying patterns, and documenting reusable insights for future contributions. structured workflow from discovery through lesson capture.

handoff captures session state and context into structured markdown documents with continuation prompts, enabling seamless resumption by the next session without losing work-in-progress state or decision context.

swarm spawns isolated parallel agents within a runtime-native backend (claude teams, codex, or fallback), respects dependency ordering and ownership constraints via pre-spawn validation, and implements fresh-context isolation per wave using filesystem persistence and messaging coordination.

security runs repeatable vulnerability and code quality checks across repositories with configurable scan depth, supporting fast pre-pr gates and full release validation modes.

bug-hunt provides a 4-phase investigation framework (root cause, pattern, hypothesis, fix) for reactive bug triage and proactive audits. outputs structured markdown reports to .agents/research/.

vibe evaluates code readiness at a slice-level quality gate by scanning complexity, hunting bugs, and applying multi-model validation to emit pass/warn/fail signals.

plan decomposes a BDD intent issue into trackable vertical slices grouped into waves, validating write scope isolation and integration order before producing markdown plans and bd issues.

forge extracts candidate learnings from session transcripts and queues them for filtering by a promotion ratchet. runs via sessionend hook to populate a learnings registry without requiring manual intervention.

pr-research guides systematic exploration of upstream repositories to understand contribution workflows and project structure before submitting pull requests to external codebases.

codex-team orchestrates parallel agent execution with lock-based file conflict prevention. agents handle discrete tasks while the lead prevents concurrent writes to the same resources.

retro captures lightweight session observations into a learnings repository, functioning as a candidate-entry surface for move 7 of an operating loop before promotion to patterns or rules.

rpi orchestrates discovery, crank, and validation phases in an autonomous workflow loop, delegating to specialized skills while enforcing invariants across the operating cycle.

beads provides git-backed task persistence using dependency graphs with triage via pagerank analysis. addresses task recovery across session boundaries where standard task lists fail.

complexity identifies refactoring candidates by analyzing code metrics and reporting hotspots. procedurally incomplete and lacks actionable recovery paths for common failure states.

doc generates and validates project documentation by detecting project type and scanning for existing docs, but lacks complete workflow definition and recovery mechanisms.

skill guides interactive creation of product.md artifacts to unlock product-aware code review perspectives. covers mission, personas, positioning, evidence, and gaps through structured interview workflow.

quickstart detects local environment setup (git, ao cli, agents directories) and provides initial orientation for new agentops users, but incomplete procedure and missing recovery paths limit utility.

update pulls the latest agentops skills from a github repository and reinstalls them globally via a curl-piped shell script. execution is incomplete.

post-mortem wraps completed work by validating correctness, extracting learnings, processing backlogs, and promoting insights into persistent memory. operates within a six-phase loop with runtime dependencies on hook surfaces.

recover detects in-progress work after context compaction and reloads relevant session state. requires codex v0.115.0+ or explicit hookless fallback via ao commands.

converter transforms agentops skills between platform formats via a three-stage parse-convert-write pipeline, but lacks actionable procedure detail and failure handling.

skill wraps openai developer docs lookups via mcp server with fallback to web search. provides product summaries but lacks actionable workflows or decision logic.

goals manages agentops goal state via yaml and markdown files using the ao goals cli, but lacks procedural detail and practical workflow documentation.

shared is a collection of reference documents and validation contracts used as dependencies by other skills. not directly invocable. minimal structural definition.

Collect independent perspectives for an explicitly high-stakes or contested judgment. Triggers: "council", "multi-judge review", "independent perspectives".

Show AgentOps work status.

Validate PR scope and quality.

Provide repo coding standards.

Plan an open source PR.

Repair skill hygiene.

Record Brownian Ratchet gates.

Run composable security analysis.

Check knowledge flywheel health.

Explore and write findings.

Trace artifact provenance.

Explain AgentOps workflows.

Reverse-engineer product specs.

Load relevant .agents context.

Implement one tracked issue.

Execute epics through waves.

Stress-test plans before work.

evolve — an installable skill for AI agents, published by boshu2/agentops.

readme — an installable skill for AI agents, published by boshu2/agentops.