agent-os

Plugin agent-os from marketplace claude-agent-os in sethdford/claude-agent-os. Commands read from the repo's manifest — 1 month ago.

Skills (18)

• ab-test

  A/B test an agent's current prompt against a candidate variant from policy/<agent>/candidates/. Runs both on the same scenarios, aggregates rewards, recommends promotion if candidate beats current by >1 stderr with n≥10. Triggers on /ab-test, "compare prompts", "test the candidate", "is the new prom

• apply-mining-patches

  Apply the diff/patch outputs from a /mine-transcripts run to lessons.md and rules/*.md after human review. Use after running /mine-transcripts and reviewing the proposed diffs. Triggers on /apply-mining-patches, "apply the mining patches", "apply mining run", "approve and apply the trajectory diffs"

• aspect-panel

  Run a panel of 5 specialized verifiers (correctness, edge-case, security, regression, style) in parallel against a change, with confidence-weighted voting. Disagreement (40-60% pass share) escalates to lead. Replaces single-critic for high-stakes review. Triggers on /aspect-panel, "panel review", "m

• best-of-n

  Run an agent N times in parallel against the same prompt, then aggregate via one of 5 modes — critic argmax, USC consistency, confidence-weighted, hybrid, or AggAgent synthesis. Use for high-stakes invocations where you'd rather pay Nx to be sure. Triggers on /best-of-n, "best of n", "run multiple",

• cache-report

  Show prompt cache hit rate trends and flag regressions. Use when user asks about cache performance, asks why sessions feel expensive, or wants to verify cache health. Triggers on /cache-report, "cache hit rate", "cache stats", "is caching working".

• caretaker

  Run the bounded harness caretaker — deterministic state gather (predictions, verifier gain, agent health, sensors, drift), then triage ONLY what's flagged under a hard authority boundary (files work, never ships changes). Triggers on /caretaker, "run the caretaker", "harness health check".

• diagnose-ci-queue

  Use when multiple PRs are stuck BLOCKED or failing CI. Clusters failures by signature, distinguishes systemic main-line bugs from per-PR content bugs, and dispatches O(root-causes) agents rather than O(PRs). Triggers on /diagnose-ci-queue, "queue is stuck", "PRs blocked", "CI is red", "merge queue j

• eval-author

  Generate scenario stubs for /eval against a target agent or skill. Reads recent successful invocations from session transcripts and proposes 5+ scenarios covering happy path, edge cases, and negative cases. Triggers on /eval-author, "author scenarios for X", "create eval scenarios", "generate test c

• eval

  Run evaluations against skills, agents, or rules. Use when the user wants to test that a skill triggers correctly, measure agent output quality, compare prompt variants, or detect regressions. Triggers on /eval, "run evals", "test skill", "benchmark agent".

• exec-grounded

  Run an agent N times in parallel against a code-change task, with each rollout in its own sandboxed copy of the codebase, scored by ACTUAL test pass rate (not just critic opinion). Use for SWE-bench-style code changes where the test suite is the ground truth. Triggers on /exec-grounded, "execution-g

• mine-transcripts

  Mine past Claude Code session transcripts (JSONL) to extract user corrections, successful patterns, and recurring failure modes. Proposes diffs against lessons.md and rules/*.md for human review. Triggers on /mine-transcripts, "mine my sessions", "what did I learn this week", "session retro". Use we

• rl-status

  Show the RL value-function snapshot for all agents — mean reward, recent trend, anomalies, top tuning candidates. Use weekly to monitor agent quality. Triggers on /rl-status, "agent rewards", "rl status", "which agents are degrading".

• scrum

  Run a complete SCRUM sprint with all ceremonies — Product Owner authors stories, Tech Lead designs, Scrum Master orchestrates implementers, Verifier+Aspect-Panel guard quality, Sprint Auditor adversarially audits, Retro feeds back to /tune-agent. Triggers on /scrum, "run a sprint", "scrum me this",

• spec

  Author a three-file spec (requirements / design / tasks) before non-trivial implementation, then verify the implementation satisfies the spec. Adopts the Kiro spec-driven development pattern. Use for features that span 3+ files, new public APIs, or anything you'd want a human to design-review. Trigg

• team

  Spin up a native Claude Code Teams fleet to tackle a multi-step task — lead orchestrates, specialists work in parallel worktrees, critic + verifier loop guards quality. Use when the task has 3+ independent sub-tasks, when you need different specialists, or when isolation between concerns is required

• tune-agent

  Propose a Reflexion-style patch to a specific agent's prompt based on evidence of recurring failures. Use when the same agent makes the same class of mistake twice, when /mine-transcripts surfaces an agent in agent-tuning.md, or when /eval flags a regression on an agent. Triggers on /tune-agent, "tu

• verify-ui

  Verify UI changes by capturing before/after screenshots and asking Claude vision to judge whether the change matches intent and didn't break anything else. Mano-verify pattern (arXiv 2509.17336 — #1 on OSWorld at 58.2%). Use after any frontend change. Triggers on /verify-ui, "verify the UI", "check

• verify

  Prove that work actually behaves correctly by spawning the verifier agent to run the code and capture evidence. Use before claiming a task complete, before commit, when in doubt about whether tests prove the right thing. Triggers on /verify, "prove it works", "verify this", "did the change actually

Subagents (26)

• accessibility-reviewer

  Use to review UI changes (HTML, JSX, Vue, Svelte, web components) against WCAG 2.1 AA — alt text, ARIA roles, keyboard navigation, color contrast, focus order, form labeling. Reports findings tagged by WCAG section with remediation and a test snippet. Read-only.

• agent-tuner

  Reflexion-style agent tuner. Given evidence of a recurring failure by a specific agent, proposes the smallest harness change (hook/tool/memory first, prompt last) that prevents the failure class. Read-write — modifies agent .md files but always backs up the prior version.

• api-contract-watcher

  Use to detect breaking changes to public APIs (HTTP endpoints, library exports, gRPC methods, CLI flags) between two git refs. Reports each break with old vs new signature and downstream impact. Read-only.

• code-reviewer
• cost-analyzer

  Use when the user wants a cloud-bill anomaly review (gcloud, AWS, Azure). Pulls current and prior period spend, identifies top deltas, and proposes hypotheses with recommendations. Read-only — does not modify resources.

• critic

  Adversarial reviewer that finds half-fixes, missing edge cases, and cross-agent regressions. Use after any task closure inside /team, before merging worktrees, or whenever you want a brutal second read. Read-only — never edits.

• dead-code-finder

  Use to find unused exports, unreachable branches, and never-imported files in a codebase. Reports findings with confidence levels because some "dead" code is dynamically loaded. Does not delete code.

• dep-auditor

  Use to audit project dependencies for known CVEs, abandoned packages, license incompatibility, and significant version drift. Reads lockfiles and consults vulnerability advisories. Does not modify lockfiles.

• devops-engineer
• doc-fleet-agent
• docs-drift-checker

  Use to detect places where code has diverged from its documentation — function signatures vs docstrings, README examples that no longer compile, OpenAPI schemas vs actual handlers, CLI help vs flags. Reports drift with suggested patches; does not edit.

• error-budget-tracker

  Use to compute SLO error budget consumption and burn rate. Reads SLO definitions and recent error/availability metrics, then reports per-SLO budget remaining and pages required if trajectory continues. Read-only.

• flake-detector

  Use when a test is suspected to be non-deterministic — passes sometimes, fails sometimes, or fails only in CI. Runs the suspect test N times under stress and reports per-test flake rate plus the suspected source. Does not modify code.

• latency-profiler

  Use to identify performance bottlenecks from profile output, slow query logs, or APM data. Reports the top hotspots with file:line references and baseline vs current p50/p95/p99 latencies. Read-only.

• migration-planner

  Use when planning a database schema migration that risks downtime, lock contention, or data loss — NULL→NOT NULL transitions, type changes, foreign key additions, large index builds, or column drops. Read-only planner that produces a staged migration plan; does NOT execute migrations.

• pipeline-agent
• product-owner

  Decomposes a user goal into a prioritized backlog of user stories with explicit, testable acceptance criteria. Use at the START of any /scrum sprint, when a goal is fuzzy, or when an existing backlog needs grooming. Read-only — never edits implementation files.

• regression-hunter

  Use when a test, behavior, or build worked at a known-good commit and is broken at HEAD. Bisects the commit history to identify the exact breaking commit and reports the failing assertion, blame, and likely fix direction. Does not edit code.

• scrum-master

  Enforces SCRUM process discipline. Sequences user stories per dependencies, dispatches work to the right specialist agents, runs daily standups, identifies blockers, ensures Definition of Done is met before story closure. Does NOT write code; orchestrates and enforces.

• security-reviewer

  Use to review code changes (a diff, branch, or file set) for OWASP Top-10 issues, hardcoded secrets, unsafe deserialization, SSRF, path traversal, and injection. Reports severity-tagged findings with remediation. Read-only.

• shell-script-specialist
• spec-verifier

  Verifies that an implementation satisfies a three-file spec (requirements / design / tasks). Reads the spec acceptance criteria and the implementation, runs evidence checks, reports per-AC PASS/FAIL with file:line refs. Distinct from the verifier agent — this one checks intent-vs-impl, not just beha

• sprint-auditor

  Adversarial end-of-sprint auditor. Independently re-reads the original user stories and the actual deliverables, then answers ONE question per AC — "did the sprint deliver this?" — without trusting the team's claims. Distinct from critic (which reviews code) and verifier (which runs code). The audit

• tech-lead

  Designs the technical approach for a sprint or specific user story. Reviews architectural decisions, identifies risks, sequences technical work. Spawned by scrum-master during planning, before implementation. Read-only — produces design docs, not implementation.

• test-specialist
• verifier

  Proves work actually behaves correctly by running it, not just checking that artifacts exist. Use immediately after any non-trivial task is claimed complete, before commit, before marking task done, and whenever the user asks to "verify that X works", "verify the function/file …", or to run tests an

Hooks (17)

• SessionStart

  ${CLAUDE_PLUGIN_ROOT}/hooks/session-start-knowledge.sh

• SessionStart

  echo '## Critical post-compaction context'; echo ''; echo 'You just compacted. Re-read your operating principles in ~/.claude/CLAUDE.md if you need the full context. Top priorities:'; echo '- /verify

• PreCompact

  echo "[$(date -u +%Y-%m-%dT%H:%M:%SZ)] Context compaction triggered" >> ~/.claude/telemetry/compactions.log

• SessionEnd

  ${CLAUDE_PLUGIN_ROOT}/hooks/log-cache-stats.sh

• SessionEnd

  python3 ${CLAUDE_PLUGIN_ROOT}/rl/emit_session_rewards.py

• SessionEnd

  python3 ${CLAUDE_PLUGIN_ROOT}/hooks/micro-mine-on-end.py

• TaskCompleted

  ${CLAUDE_PLUGIN_ROOT}/hooks/auto-verify-on-complete.sh

• TaskCompleted

  python3 ${CLAUDE_PLUGIN_ROOT}/rl/emit_task_reward.py

• UserPromptSubmit

  python3 ${CLAUDE_PLUGIN_ROOT}/rl/emit_correction_signal.py

• UserPromptSubmit

  ${CLAUDE_PLUGIN_ROOT}/hooks/completion-claim-detector.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/constitutional-gate.sh

• PreToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/auto-critic-on-commit.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/auto-eval-on-agent-edit.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/integration-done-audit.sh

• PostToolUse

  ${CLAUDE_PLUGIN_ROOT}/hooks/agent-crash-capture.sh

• Stop

  ${CLAUDE_PLUGIN_ROOT}/hooks/stop-verify-gate.sh

• TeammateIdle

  ${CLAUDE_PLUGIN_ROOT}/hooks/teammate-idle.sh

Similar plugins