dev
Skill1 starsChanged 5 months agoApache-2.0View on GitHub
E2E development: investigate → dig → decompose → implement → test → review → PR
- Commits and pushes
56 other projects use a version of this.
Adopted by
Most recently taken up by
- TsukiNi22/skillstheir own version, first seen Nov 16, 2026
- TsukiNi22/ai-utilstheir own version, first seen Nov 16, 2026
- waynesutton/convexskillstheir own version, first seen Nov 15, 2026
- waynesutton/builder-skillstheir own version, first seen Nov 15, 2026
- ztcools/skillstheir own version, first seen Nov 15, 2026
- mayoz/rabbittheir own version, first seen Nov 15, 2026
Used with
Projects with it also use metrics-graphana, control-flow and 10 more.
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.
slopusSkillmetrics-graphana
Query and manage Grafana dashboards and Prometheus metrics for Happy infrastructure. Covers grafanactl CLI usage, direct Prometheus queries through Grafana proxy, and dashboard-as-code workflows. Use when user asks about metrics, dashboards, monitoring, Grafana, Prometheus, or wants to add/modify panels.
Also in 3 projects that use this
slopusSkillcontrol-flow
Analyze and design control flows and data structures. Produces compact ASCII tree diagrams showing triggers, call chains, payload shapes, state mutations, and re-render effects. Use when user asks to diagram, trace, visualize, or design a flow or data structure.
Also in 3 projects that use this
waynesuttonSkillsec-check
Security review checklist for Convex functions, auth logic, public queries, admin routes, webhooks, uploads, and AI-generated code. Use when reviewing code that touches user data, PII, or access control.
Also in 3 projects that use this
waynesuttonSkillupdate-project-docs
Also in 3 projects that use this
silverkyoon-RachelSkillstyle
Tailwind CSS 스타일을 검토하고 개선합니다.
Also in 3 projects that use this
jasonmichaelbell78-creatorSkillaudit-security
Run a single-session security audit on the codebase
Also in 3 projects that use this
Other versions56 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 56
- slopus/happy★ 25ksame name
- microsoft/playwright-cli★ 15ksame name
- atopile/atopile★ 4ksame name
- cexll/myclaude★ 2.8ksame name
- waynesutton/markdown-site★ 628same name
- Automattic/wordpress-activitypub★ 583same name
- waynesutton/builder-skills★ 404same name
- waynesutton/convexskills★ 404same name
- justrach/merjs★ 357same name
- stella/stella★ 280same name
46 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 repository12 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- auditSkill
- decomposeSkill
- dev-allSkill
- dev-investigateSkill
- digSkill
- investigateSkill
- prSkill
- reviewSkill
- tech-debtSkill
- update-docsSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
32 rules. 0 are common, 17 are distinctive, and 15 are very specific to this project.
0
Common
Most of the 38 edited versions we read keep these.
17
Distinctive
This author's own choices. Worth a look.
15
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 42
Use TaskCreate to create a task for each phase.
- Very specific: line 54
Use TaskUpdate to mark each task in_progress when starting and completed when done.
- Very specific: line 95
Use ToolSearch with +linear to load the Linear MCP
- Very specific: line 96
Call mcp__linear__get_issue with the issue ID
- Very specific: line 99
Use ToolSearch with +figma-remote to load Figma MCP
- Distinctive: line 106
Read CLAUDE.md for project architecture and conventions.
- Distinctive: line 110
If the issue involves library/framework updates, version bumps, or API deprecations:
- Very specific: line 111
Check labels for migration, upgrade, dependency, or breaking-change
- Very specific: line 112
If Android/KMP project and android CLI is available: run android docs search "{migration topic}" to get latest official migration guidance
- Distinctive: line 129
After the skill completes, read the investigation report:
- Distinctive: line 137
After reading the report:
- Very specific: line 142
If anything is ambiguous, use AskUserQuestion.
- Distinctive: line 152
Use Codex to generate a technical design before diving into ambiguity resolution and decomposition.
- Distinctive: line 162
If Codex is available:
- Very specific: line 191
If ToolSearch fails to find Codex or the call errors:
- Very specific: line 203
Use the /dig skill with investigation results (and Codex design if available) to resolve decision points.
- Very specific: line 213
Use the /decompose skill to break the work into ordered subtasks.
- Distinctive: line 226
Ask the user to confirm before implementation.
- Distinctive: line 270
Follow existing code patterns (read surrounding code first)
- Distinctive: line 272
Keep changes minimal and focused
- Distinctive: line 282
Run the project's build, test, and lint commands as defined in CLAUDE.md's Commands section.
- Distinctive: line 284
If CLAUDE.md doesn't specify commands, detect from project files:
- Very specific: line 304
Use the /review skill to run multi-agent parallel review.
- Very specific: line 308
After the review completes, write workspace/{issue}/review.json:
- Distinctive: line 335
Critical: STOP. Report to user. Do NOT proceed.
- Distinctive: line 363
No Co-Authored-By, no AI stamps
- Very specific: line 370
Use the project's pull_request_template.md if available.
- Very specific: line 391
When the user invokes /dev with /goal, the workflow runs autonomously:
- Distinctive: line 399
Stop on Critical review findings or 3 consecutive failures (these still require human input)
- Distinctive: line 427
Critical review findings: Always stop and report
- Distinctive: line 428
3 consecutive failures: Always stop and report
- Distinctive: line 429
Unexpected errors: Always stop and report
“Common” is measured: kept by at least 6 in 10 of the 38 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
/dev — E2E Development Workflow
Resolve Issue $ARGUMENTS from investigation to PR creation.
Target: $ARGUMENTS
Setup: Create Task Tracker
Use TaskCreate to create a task for each phase. This provides progress visibility and persistence across /compact.
- "Gather context from issue"
- "Investigate codebase (/dev-investigate)"
- "Technical design (Codex)"
- "Resolve ambiguities (/dig)"
- "Decompose into subtasks (/decompose)"
- "Implement changes"
- "Run quality gate"
- "Review changes"
- "Commit & create PR"
Use TaskUpdate to mark each task in_progress when starting and completed when done.
Workflow
Phase 1: Issue Understanding
↓
Phase 2: Investigation (← /dev-investigate, context: fork)
↓
Phase 2.5: Technical Design (← Codex, optional)
↓
Phase 3: Ambiguity Resolution (/dig)
↓
Phase 4: Task Decomposition (/decompose)
↓
── AskUserQuestion: confirm approach + task list ──
↓
Phase 5: Branch & Implement
↓
Phase 6: Quality Gate (build + test + lint from CLAUDE.md)
↓
Phase 7: Review (/review)
↓
── AskUserQuestion: commit + PR confirmation ──
↓
Phase 8: Commit & PR CreationPhase 1: Issue Understanding
Mark task 1 in_progress.
Detect the issue source from "$ARGUMENTS":
GitHub Issue (starts with # or is a number):
gh issue view <number> --json number,title,body,labels,assignees,comments- Extract: title, description, acceptance criteria, labels
Linear Issue (matches XXX-1234 pattern):
- Use
ToolSearchwith+linearto load the Linear MCP - Call
mcp__linear__get_issuewith the issue ID
Figma links in the issue description (figma.com/design/...):
- Use
ToolSearchwith+figma-remoteto load Figma MCP - Fetch design context and screenshot
Branch naming (from labels or issue type):
- Bug →
fix/{issue-ref}-{kebab-case-short-desc} - Otherwise →
feat/{issue-ref}-{kebab-case-short-desc}
Read CLAUDE.md for project architecture and conventions.
Migration Detection
If the issue involves library/framework updates, version bumps, or API deprecations:
- Check labels for
migration,upgrade,dependency, orbreaking-change - If Android/KMP project and
androidCLI is available: runandroid docs search "{migration topic}"to get latest official migration guidance - Note migration-specific risks (breaking changes, deprecated APIs) for Phase 3 (/dig)
Mark task 1 completed.
Phase 2: Investigation (/dev-investigate)
Mark task 2 in_progress.
Invoke the /dev-investigate skill to run investigation in a forked context. This keeps large Read results isolated from this context.
Skill("dev-investigate", args: "{issue title}. {issue description}. Keywords: {keywords}. Affected areas: {areas from labels/description}")After the skill completes, read the investigation report:
Read("investigation-report.md")Think Twice
After reading the report:
- Is the investigation thorough? All affected files identified?
- Are there other possible causes not considered?
- Is impact analysis complete?
If anything is ambiguous, use AskUserQuestion. Never assume.
Mark task 2 completed.
Phase 2.5: Technical Design (Codex)
Mark task 3 in_progress.
Use Codex to generate a technical design before diving into ambiguity resolution and decomposition.
Load Codex
ToolSearch("select:mcp__codex__codex")Call Codex
If Codex is available:
mcp__codex__codex(
prompt: "Given this investigation report and issue, produce a technical design:
## Issue
{issue title and description}
## Investigation Findings
{summary from investigation-report.md}
## Affected Files
{affected files table from report}
## Design Request
1. Propose the implementation approach (architecture, data flow changes)
2. Identify interface changes and new abstractions needed
3. List edge cases and error handling strategy
4. Flag risks and trade-offs
Output a concise technical design document."
)Save the design output for use in Phase 3 (/dig) and Phase 4 (/decompose).
Fallback (Codex unavailable)
If ToolSearch fails to find Codex or the call errors:
- Skip this phase — proceed to Phase 3 with investigation results only (traditional flow)
- Log: "Codex unavailable, skipping technical design phase"
Mark task 3 completed.
Phase 3: Ambiguity Resolution
Mark task 4 in_progress.
Use the /dig skill with investigation results (and Codex design if available) to resolve decision points.
Mark task 4 completed.
Phase 4: Task Decomposition
Mark task 5 in_progress.
Use the /decompose skill to break the work into ordered subtasks.
Mark task 5 completed.
── AskUserQuestion: Approach Confirmation ──
Present to the user:
- Decision Matrix (from /dig)
- Task List (from /decompose, with dependencies)
- Investigation summary (key findings)
Ask the user to confirm before implementation.
Phase 5: Branch & Implement
Mark task 6 in_progress.
5a. Create Branch
git checkout -b {branch-name}5b. Implement
TDD mode (when issue has tdd label, or test changes are the primary goal):
LOOP for each subtask:
1. TaskUpdate → in_progress
2. Read target code
3. Write/update tests FIRST (use test-writer agent if needed)
4. Run tests — confirm they FAIL (red)
5. Implement the minimal code to pass
6. Run tests — confirm they PASS (green)
7. Refactor if needed (keep tests passing)
8. TaskUpdate → completedStandard mode (default):
LOOP for each subtask (in dependency order):
1. TaskUpdate → in_progress
2. Read target code (MUST read before editing)
3. Implement changes (Edit/Write)
4. Self-verify (run Verify step from task description)
5. TaskUpdate → completed
INTERRUPT conditions:
- Unexpected problem → AskUserQuestion
- 3 consecutive failures → STOP and reportGuidelines:
- Follow existing code patterns (read surrounding code first)
- Follow CLAUDE.md conventions
- Keep changes minimal and focused
Mark task 6 completed.
Phase 6: Quality Gate
Mark task 7 in_progress.
Run the project's build, test, and lint commands as defined in CLAUDE.md's Commands section.
If CLAUDE.md doesn't specify commands, detect from project files:
build.gradle.kts/gradlew→./gradlew build,./gradlew test,./gradlew detektpackage.json→npm test,npm run lintCargo.toml→cargo build,cargo test,cargo clippypyproject.toml/setup.py→pytest,ruff check
Failure Handling
- Analyze the failure
- Fix the issue
- Re-run the failing check
- Maximum 3 fix attempts — if still failing, report to user and stop
Mark task 7 completed.
Phase 7: Review
Mark task 8 in_progress.
Use the /review skill to run multi-agent parallel review.
Structured Review Output
After the review completes, write workspace/{issue}/review.json:
{
"issue": "{issue reference}",
"branch": "{branch name}",
"timestamp": "{ISO 8601}",
"status": "clean | warnings | critical",
"counts": {
"critical": 0,
"warning": 0,
"suggestion": 0,
"nit": 0
},
"findings": [
{
"severity": "critical | warning | suggestion | nit",
"file": "path/to/file",
"line": 42,
"description": "description of finding",
"resolved": false
}
]
}Review Result Handling
- Critical: STOP. Report to user. Do NOT proceed.
- Warning: Fix, re-run Quality Gate (Phase 6). Update
review.jsonfindings as"resolved": true. - Suggestion: Note but don't block
Mark task 8 completed.
── AskUserQuestion: Commit + PR Confirmation ──
Show the user:
- Summary of all changes
- Quality gate results
- Review findings and resolutions
- Proposed commit message (single line, no AI stamps)
Phase 8: Commit & PR Creation
Mark task 9 in_progress.
8a. Commit
git add {specific files}
git commit -m "{concise message}"- Explicit file staging (no
git add .) - No Co-Authored-By, no AI stamps
8b. Push & PR
git push -u origin {branch-name}Use the project's pull_request_template.md if available. Only fill in Description and Related Issues.
gh pr create --title "#{issue} {description}" --body "$(cat <<'EOF'
## Description
- {bullet point summary}
## Related Issues
Closes #{issue}
EOF
)"Report PR URL to the user.
Mark task 9 completed.
Autonomous Mode (/goal)
When the user invokes /dev with /goal, the workflow runs autonomously:
/goal "Issue $ARGUMENTS is resolved: tests pass, review has no Critical findings, and PR is created"In autonomous mode:
- Skip
AskUserQuestionconfirmations — proceed with best judgment - Stop on Critical review findings or 3 consecutive failures (these still require human input)
- The
/goalevaluator checks the completion condition after each phase
Structured Return Value
On completion, output a structured result for callers (e.g., /dev-all):
{
"issue": "{issue reference}",
"status": "success | failed | blocked",
"pr_url": "https://github.com/owner/repo/pull/N",
"review": {
"status": "clean | warnings | critical",
"critical_count": 0,
"warning_count": 0
},
"failure_reason": null
}AskUserQuestion Skip Rules
In autonomous mode, AskUserQuestion is skipped when:
- Approach confirmation (Phase 4): Proceed with the decomposed task list
- Commit + PR confirmation (Phase 8): Proceed if review status is
cleanorwarnings(no unresolved criticals)
AskUserQuestion is NOT skipped when:
- Critical review findings: Always stop and report
- 3 consecutive failures: Always stop and report
- Unexpected errors: Always stop and report
Error Handling
| Situation | Action |
| Issue not found | Report error, stop |
| Figma fetch fails | Warn, continue without design |
| Investigation unclear | AskUserQuestion before proceeding |
| Tests fail (≤3 attempts) | Fix and retry |
| Tests fail (>3 attempts) | Report to user, stop |
| Critical review finding | Report to user, stop |
| Warning review finding | Fix, re-run quality gate |
| Git/PR creation fails | Report error, stop |
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.

