start-task
Skill2 starsChanged 6 days agoMITView on GitHub
Start a task — load the beads context, interview the user before any code is touched, create and claim a bead, read the affected files, propose a plan, and wait for an explicit go-ahead. Use for /start-task, or when the user wants to begin a new piece of work and no bead is claimed yet.
18 other projects use a version of this.
Adopted by
Most recently taken up by
- seancabalse/agentic-playbooktheir own version, first seen Nov 16, 2026
- Krzysztof318/MailFathomtheir own version, first seen Nov 15, 2026
- maotoumao/Cebiantheir own version, first seen Nov 9, 2026
- BSSE23047/nauraiz-devtheir own version, first seen Nov 8, 2026
- paologum/claude-skillstheir own version, first seen Oct 25, 2026
- a-ghorbani/pocketpal-dev-teamtheir own version, first seen Oct 18, 2026
Used with
Projects with it also use review.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
SentrySkillreview
Fixture review skill.
Also in 3 projects that use this
Other versions18 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 18
- devoxx/DevoxxGenieIDEAPlugin★ 689same name
- nockawa/Typhon★ 253same name
- mattias800/prosper★ 217same name
- maotoumao/Cebian★ 185same name
- fathomise/MailFathom★ 21same name
- Krzysztof318/MailFathom★ 21same name
- ymatrix1/vantage★ 17same name
- a-ghorbani/pocketpal-dev-team★ 6same name
- sensei-hacker/inav-claude★ 6same name
- lutzleonhardt/casefile★ 4same name
8 more not listed
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 repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- changelogSkill
- close-shopSkill
- deferSkill
- finalizeSkill
- review-titlesSkill
- statusSkill
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 8 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 16 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 9
Run bd prime to load complete workflow context
- Very specific: line 10
Run bd ready to show available work
- Very specific: line 11
If a bead ID is provided as an argument, claim it: bd update <id> --claim and skip to step 3.
- Distinctive: line 14
If no bead ID was given, interview the user first:
- Distinctive: line 15
Ask what they want to do and why
- Distinctive: line 32
If neither exists, skip this step.
- Distinctive: line 36
Use Glob and Grep to locate relevant files
- Distinctive: line 39
Read the identified files to understand existing patterns
- Distinctive: line 40
Do not skim -- understand before proposing
- Distinctive: line 47
Ask any remaining questions
- Very specific: line 55
Update the bead with any key design decisions, architectural choices, or non-obvious tradeoffs that emerged during the discussion (bd …
- Distinctive: line 57
Do this BEFORE writing code, so the decisions are captured even if the session is interrupted
“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
This skill changes no code. It ends with a plan, and implementation waits for the user's go-ahead.
1. Load Context
- Run
bd primeto load complete workflow context - Run
bd readyto show available work - If a bead ID is provided as an argument, claim it:
bd update <id> --claimand skip to step 3.
2. Create a Bead (if none provided)
If no bead ID was given, interview the user first:
- Ask what they want to do and why
- Clarify scope, affected areas, and constraints
- Once you understand the task, create a bead with
bd createand claim it - Then continue to step 3.
3. Interview the User (DISCUSS ONLY)
Ask about:
- Task scope and goals
- Which parts of the system are affected
- Specific requirements and constraints
- Any related issues or dependencies
4. User- and Project-Specific Context
Extensions are read from two places, user first, then project, and both are followed:
~/.claude/skill-extensions/start-task.md— the user's own conventions, the same in every project..claude/skill-extensions/start-task.mdin the project — that project's.
If neither exists, skip this step.
5. Identify Source Files
- Find all components, endpoints, models, migrations, and tests related to the task
- Use Glob and Grep to locate relevant files
6. Read Source Files
- Read the identified files to understand existing patterns
- Do not skim -- understand before proposing
- State your key assumptions about how the existing code works (data flow, state management, naming conventions) so the user can catch misunderstandings early
7. Summarize and Discuss
- Include the bead ID (e.g.,
[proj-abc123]) in your summary for traceability - Confirm your understanding of the task
- Describe what files will be changed and why
- Ask any remaining questions
- Propose your implementation plan
- Suggest a session name: end the summary by proposing a short, meaningful name (a few words) that captures this task, and remind the user to apply it with
/rename <name>. Renaming is a user-only action — no agent tool can do it — so the user must run the command themselves.
Implementation begins only after the user says "go ahead", "implement", "do it", or equivalent.
8. Record Design Decisions
Once discussion is complete and the user has given permission to implement:
- Update the bead with any key design decisions, architectural choices, or non-obvious tradeoffs that emerged during the discussion (
bd update <id> --append-notes="...") - This preserves context for future sessions, code reviews, and commit authoring
- Do this BEFORE writing code, so the decisions are captured even if the session is interrupted
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.

