project-manager
Runs resumable long-horizon builds by decomposing dependency-aware slices, dispatching isolated workers, reviewing and integrating checkpoints, and verifying…
Skill by 0xdewy
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/0xdewy-project-manager
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/project-manager
What's inside
- Skill name
project-manager- Description
- Runs resumable long-horizon builds by decomposing dependency-aware slices, dispatching isolated workers, reviewing and integrating checkpoints, and verifying the result; supports explicit competing candidates. Only use when explicitly requested.
- License
- MIT
- Instructions
- 56 lines in SKILL.md · view on GitHub
Files (3)
- SKILL.md
- evals.json · evals/
- subagent-prompts.md · references/
SKILL.md (first lines)
# Project Manager Own a long-running mandate from decomposition through verified integration. Load `../common/patterns/execution-contract.md` and `../common/patterns/scaling.md`. Load `../common/patterns/worker-slice.md` and `references/subagent-prompts.md` only when dispatching or reviewing workers. Validate state with `../common/scripts/validate_state.py`. ## Modes - `--quick`: direct work or one worker; one review checkpoint. - `--standard`: 2-4 dependency-aware slices; one revision per failing slice. - `--thorough`: multiple bounded waves with explicit worker/wave budget. - `--compete`: orthogonal flag; dispatch 2-3 isolated candidates against the same acceptance criteria, test all, and select by evidence. Use this only when workstreams are independent or the task must persist across sessions. A single verifier loop is `goal`; one direct action needs no skill. ## Durable State Use the emitted schema (`validate_state.py project-manager --print-template`) for the single-writer `status.json`. Record mandate, mode, state, current wave, acceptance criteria, integration checks, `next_action`, and each slice's ID, dependencies, owner, state, attempt count, artifact, and evidence. Slice states: `pending`, `ready`, `in_progress`, `review`, `accepted`, `blocked`. If an unfinished workspace matches the mandate, validate artifacts and resume from `next_action`; never redo accepted slices. Archive only completed or mismatched runs. Checkpoint after planning, each wave, integration, and final verification so interruption loses no accepted work. ## Workflow 1. State mode/workspace and create `PLAN.md` plus validated `status.json`. Define slices, dependencies, owned paths, acceptance, integration order, and checks. 2. Mark dependency-satisfied slices `ready`; dispatch only disjoint ready work in parallel under `workers/<id>/`. 3. Validate each worker artifact, inspect its diff, run slice checks, and record `ACCEPT` or one targeted revision. Never accept conversational claims alone. 4. Advance the dependency graph; validate and checkpoint after every wave. 5. Integrate accepted slices in declared order. Preserve user-owned worktree changes and resolve cross-slice conflicts centrally. 6. Under `--compete`, keep candidates isolated and reject any not run through the same checks; document the selection and discarded tradeoffs. 7. Run integration checks. Optional `red-team` is report-only and only when explicitly requested; map accepted findings back to criteria before fixes. 8. Write final evidence and `next_action: null`. Never commit unless separately requested. ```text DONE: <WORKSPACE> — <accepted>/<total> slices accepted, waves=<N>, verification=<passed|partial|blocked> ```
From 0xdewy/skills, MIT.
More from 0xdewy/skills
- 0.570score
agentify
0xdewy
Maps git-tracked code with a ranked outline, scoped expansion, and symbol or call-site XREFs.
SkillAFreshness: AClaude Code +9 - 0.570score
audit-debian-server
0xdewy
Audit, harden, and monitor Debian or Ubuntu application servers.
SkillAFreshness: AClaude Code +9 - 0.570score
code-smellz
0xdewy
Finds and fixes repository-wide correctness, security, architecture, and maintainability problems with bounded review.
SkillAFreshness: AClaude Code +9 - 0.570score
frontend-qa
0xdewy
Drives a local web app in a headless browser to find and fix functional failures and visual defects, then proves each fix with a re-run.
SkillAFreshness: AClaude Code +9 - 0.570score
git-show
0xdewy
Visualize repository fleets, PR stacks, and local work in an interactive HTML dashboard or compact text tree.
SkillAFreshness: AClaude Code +9 - 0.570score
shrinkray
0xdewy
Shrinks repositories while preserving behavior by removing dead code, duplication, and verbosity with measured verification.
SkillAFreshness: AClaude Code +9 - 0.570score
spec-reconciler
0xdewy
Reconciles source-anchored claims across schemas, code, tests, docs, and generated clients using declared authority.
SkillAFreshness: AClaude Code +9 - 0.570score
student-counsel
0xdewy
A Student does the work and Socrates examines it by elenchus until the work can answer every question without contradiction, or the round cap lands.
SkillAFreshness: AClaude Code +9 - 0.570score
vfs-docs
0xdewy
Builds self-describing docs/ trees where directory shape and filenames carry the meaning, so an agent orients from `tree docs/` alone.
SkillAFreshness: AClaude Code +9 - 0.570score
invariant-miner
0xdewy
Mines evidence-backed invariants and falsifies them with generative tests, shrinking reproducible counterexamples.
SkillAFreshness: AClaude Code +9 - 0.570score
llm-providers
0xdewy
Developer reference for connecting to OpenAI, Anthropic, Gemini, DeepSeek, GLM, MiniMax, MiMo, Mistral, xAI, Qwen, Kimi, Cohere, OpenRouter, Together, and…
SkillAFreshness: AClaude Code +9 - 0.570score
pr-describe
0xdewy
Writes or rewrites a pull request's title and description so a reader with no context understands why, what changed, and how to verify it.
SkillAFreshness: AClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 2
- Last commit (this folder)
- Oct 11, 2026 (17 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/0xdewy/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from 0xdewy
- 0.570score
audit-debian-server
0xdewy
Audit, harden, and monitor Debian or Ubuntu application servers.
SkillAFreshness: AClaude Code +9 - 0.570score
agentify
0xdewy
Maps git-tracked code with a ranked outline, scoped expansion, and symbol or call-site XREFs.
SkillAFreshness: AClaude Code +9 - 0.570score
frontend-qa
0xdewy
Drives a local web app in a headless browser to find and fix functional failures and visual defects, then proves each fix with a re-run.
SkillAFreshness: AClaude Code +9 - 0.570score
git-show
0xdewy
Visualize repository fleets, PR stacks, and local work in an interactive HTML dashboard or compact text tree.
SkillAFreshness: AClaude Code +9 - 0.570score
code-smellz
0xdewy
Finds and fixes repository-wide correctness, security, architecture, and maintainability problems with bounded review.
SkillAFreshness: AClaude Code +9 - 0.570score
invariant-miner
0xdewy
Mines evidence-backed invariants and falsifies them with generative tests, shrinking reproducible counterexamples.
SkillAFreshness: AClaude Code +9
Cite this page
project-manager skill. Agent Mart. https://agentmart.dev/skills/0xdewy-project-manager (last updated 2026-10-24)