big-task
Skill4 starsChanged 54 days agoMITView on GitHub
Coordinate an engineering workflow when the user explicitly invokes big-task or requests an end-to-end engineering cycle. Choose planning, implementation, review, and verification depth from actual risk and stop at the requested deliverable. Ordinary features, file counts, and refactors do not trigger this lifecycle by themselves.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- lippdev/claude-skilltheir own version, first seen Oct 11, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- lippdev/claude-skill★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository39 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agentic-repo-scaffoldSkill
- agent-travel-plannerSkill
- ai-workflow-spineSkill
- ana-awardSkill
- analyze-video-evidenceSkill
- apple-pdfSkill
- audit-fix-loopSkill
- autonomous-grindSkill
- award-flightsSkill
- banxian-skillSkill
- clean-room-rewriteSkill
- cloudflare-r2-setupSkill
- codebase-sweepSkill
- context-infrastructureSkill
- deep-researchSkill
- download-web-videoSkill
- dr-sharpSkill
- game-script-creationSkill
- gemini-imageSkill
- gpt-imageSkill
- investigateSkill
- jewelry-marketingSkill
- learning-coachSkill
- mineru-pdf-parserSkill
- mtcSkill
- narrative-researchSkill
- neat-freakSkill
- pet-forgeSkill
- plan-design-reviewSkill
- pr-fix-loopSkill
- proxy-node-setupSkill
- serenity-bottleneck-researchSkill
- task-shaped-agent-routingSkill
- transcribeSkill
- tridentSkill
- tui-engineeringSkill
- wandian-writerSkill
- web-novel-writingSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Read the user's request and standing project instructions together.
- Distinctive: line 43
Use profile-detection for unfamiliar repos and routing-examples for ambiguous work shapes.
- Distinctive: line 53
Use bounded parallel agents when independent work can improve speed or quality and the environment permits delegation.
- Distinctive: line 58
Read returned evidence and integrate the result.
- Distinctive: line 61
Run checks that exercise the changed behavior and required repository contracts.
- Distinctive: line 69
Use review-discipline for consequential review questions and adaptation-hints for migrations, UI references, or AI behavior.
- Distinctive: line 77
All acceptance criteria must be implemented and supported by evidence, or identified as unresolved with the concrete blocker.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Big Task
Carry the requested engineering outcome through implementation and verification, using the lightest workflow that protects it.
Scope and authority
Read the user's request and standing project instructions together. Preserve accepted phases, acceptance criteria, existing authorization, and user-owned changes. A review or planning request remains read-only unless edits are requested. Invoking this router does not itself request commits, push, PRs, remote messages, merge, deployment, memory promotion, or repository cleanup.
Continue authorized phases without asking again at each boundary. Resolve routine reversible choices using the repository's patterns. When a material choice cannot be inferred, prepare the concrete options and recommendation, ask once for the missing decision, and continue independent authorized work. Respect explicit sign-off gates; elapsed time or silence is not approval.
Establish the working contract
Inspect the applicable repository instructions, manifests, relevant source, tests, and supplied references. Identify the actual source of truth before editing generated files or copies. Keep the outcome, acceptance criteria, scope, verification, and material unknowns visible in a short plan. Reuse an existing plan; persist a new one only when dependencies or continuity need it.
Choose depth from ambiguity, reversibility, blast radius, and user/data impact:
| Work shape | Useful approach |
| Clear change following an established pattern | Implement directly and run focused checks. |
| Several dependent deliverables | Keep an acceptance checklist and execute dependency order. |
| Consequential architecture, persistence, concurrency, or public contract change | Resolve invariants and compatibility first; validate the risky path with focused evidence and specialist review when it adds confidence. |
| Unproven feasibility or unclear product direction | Investigate the consequential uncertainty before expensive implementation. |
File count, language, and the presence of a database do not determine process. Use profile-detection for unfamiliar repos and routing-examples for ambiguous work shapes.
Execute and verify
Own the bounded implementation end to end. Follow established patterns, make the requested behavior work, and update directly affected documentation when the change makes it inaccurate. Keep unrelated refactors and housekeeping out of the diff.
Use bounded parallel agents when independent work can improve speed or quality and the environment permits delegation. Give each task explicit inputs, acceptance criteria, and file ownership. Keep coupled changes sequential; use isolated worktrees when shared state would conflict. Choose available capabilities and models from the actual task, without fixed role counts. Read returned evidence and integrate the result; delegation does not replace ownership or require another generic reviewer.
Run checks that exercise the changed behavior and required repository contracts. Add regression tests where they protect a demonstrated failure mode. For visual changes, inspect rendered output against the relevant design reference; choose screens and states affected by the change. Reuse passing evidence until the diff or unresolved risk changes. Do not introduce coverage targets, test scaffolding, or repeated full-suite runs merely to satisfy this workflow.
Use review-discipline for consequential review questions and adaptation-hints for migrations, UI references, or AI behavior. A broad audit or PR-fix loop runs only when it is part of the user's requested workflow. Each keeps its own scope and completion contract.
Finish at the requested outcome
All acceptance criteria must be implemented and supported by evidence, or identified as unresolved with the concrete blocker. A phase boundary is not a reason to stop while authorized work remains. A successful local build is not proof of deployment or PR approval.
Honor authorized commit, push, PR, merge, and deployment phases when they are in scope; do not add them as a default ending. Use repository conventions for commit granularity and PR format. Remote review requests or bot messages need explicit authorization in the session.
Report what changed, verification results, and material limitations with links to the deliverable. Use report-templates only if a reporting format is useful. For explicitly requested cleanup, the cleanup reference provides removal checks; cleanup is not a final phase of ordinary implementation.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

