update-claude-documentation
Skill8 starsChanged 14 days agoMITView on GitHub
Use when repository behaviour, routing, counts, validation, or release procedures change and README, AGENTS, CLAUDE, CONTRIBUTING, or focused docs must be synchronised; not for ordinary copy edits.
In this repository101 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/profiles-sectors/sectors/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- ai-agent-commercial-packagingSkill
- ai-agent-contract-language-packSkill
- ai-agent-intervention-credit-and-abort-refundSkill
- ai-agent-msa-and-sla-addendum-templatesSkill
- ai-agent-procurement-objections-on-commercialsSkill
- ai-agent-renewal-and-true-upSkill
- ai-agent-sla-and-credit-scheduleSkill
- ai-agent-success-fee-and-outcome-pricingSkill
- ai-agent-business-case-and-roiSkill
- ai-agent-change-management-and-adoptionSkill
- ai-agent-compliance-credentialsSkill
- ai-agent-discovery-and-qualificationSkill
- ai-agent-methodologySkill
- ai-agent-poc-and-pilot-scopingSkill
- ai-agent-pricing-and-packaging-proposalSkill
- ai-agent-procurement-and-questionnaireSkill
- ai-agent-risk-and-responsible-aiSkill
- ai-agent-team-compositionSkill
- ai-agent-vertical-positioningSkill
- ai-on-saas-business-case-and-roiSkill
- ai-on-saas-change-management-and-adoptionSkill
- ai-on-saas-combined-methodologySkill
- ai-on-saas-compliance-credentialsSkill
- ai-on-saas-discovery-and-qualificationSkill
- ai-on-saas-poc-and-pilot-scopingSkill
- ai-on-saas-pricing-and-packaging-proposalSkill
- ai-on-saas-procurement-and-questionnaireSkill
- ai-on-saas-risk-and-responsible-aiSkill
- ai-on-saas-team-compositionSkill
- ai-on-saas-vertical-positioningSkill
- accounting-finance-advisorySkill
- business-analysis-toolsSkill
- capacity-buildingSkill
- change-managementSkill
- consulting-frameworksSkill
- data-managementSkill
- eac-ecommerce-bds-programme-designSkill
- environmental-and-social-safeguardsSkill
- gender-and-social-inclusionSkill
- giz-eu-local-procurement-responseSkill
- monitoring-and-evaluationSkill
- project-managementSkill
- retail-transformation-proposalSkill
- risk-managementSkill
- stakeholder-engagementSkill
- sustainability-planningSkill
- east-african-englishSkill
- language-standardsSkill
- ai-slop-auditSkill
- anti-ai-slopSkill
- bid-red-team-dual-reviewSkill
- kaizen-improvement-systemSkill
- operational-readiness-and-localisationSkill
- skill-safety-auditSkill
- skill-writingSkill
- submission-proof-and-receipt-disciplineSkill
- 01-cover-letterSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 13 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈13
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
Use this skill when repository documentation such as README.md, CLAUDE.md, or repo docs must be updated after skill changes.
- Distinctive: line 32
Update specific technical or contract documentation before general summaries.
- Distinctive: line 33
Stop if machine output and a documented count or command disagree.
- Distinctive: line 35
Verify links, paths, commands, counts, and release instructions across the affected documents.
- Distinctive: line 38
Keep documentation factual, current, and aligned to actual file paths and workflows.
- Distinctive: line 39
Prefer minimal updates that eliminate drift without adding unnecessary ceremony.
- Distinctive: line 63
Read and search are required.
- Distinctive: line 82
Update project documentation systematically after significant changes.
- Very specific: line 86
Modularize Instructions (Token Economy): Avoid packing everything into a single CLAUDE.md.
- Distinctive: line 171
Read all affected files in parallel.
- Distinctive: line 184
Check across all files:
- Distinctive: line 268
Update specific docs first, general last
- Distinctive: line 269
Read all affected files before editing
- Distinctive: line 272
Test examples before committing
≈ “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
Update Claude Documentation
Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.
<!-- dual-compat-start -->Use When
- Use this skill when repository documentation such as README.md, CLAUDE.md, or repo docs must be updated after skill changes.
- Load it when repository behavior has changed and the Claude-facing docs are now incomplete or stale.
Do Not Use When
- The task does not change repository behavior or documentation-relevant structure.
- The work is unrelated to repository documentation.
Inputs
| Artefact | Source/provider | Required? | Missing-input behaviour |
| Implemented change and complete diff | Repository | Yes | Stop and inspect the change before describing it. |
| Affected documentation | Repository routers and docs | Yes | Mark inaccessible documents not assessed; do not claim consistency. |
Workflow
- Review the implemented change and map facts to affected documents.
- Update specific technical or contract documentation before general summaries.
- Stop if machine output and a documented count or command disagree.
- Recover by rerunning the source check, correcting stale claims, and reconciling terminology.
- Verify links, paths, commands, counts, and release instructions across the affected documents.
Quality Standards
- Keep documentation factual, current, and aligned to actual file paths and workflows.
- Prefer minimal updates that eliminate drift without adding unnecessary ceremony.
- Preserve compatibility with existing repository workflows and file paths.
Anti-Patterns
- Updating README while leaving AGENTS routing stale. Fix: map the change to every governing document.
- Describing a validator that was not run. Fix: quote only commands and results supported by execution evidence.
- Copying the same long procedure into three routers. Fix: keep the procedure in one focused document and link to it.
- Preserving a stale skill count after filesystem changes. Fix: derive counts from the live catalogue.
- Claiming a breaking change is harmless. Fix: state the migration or compatibility consequence explicitly.
Outputs
| Artefact | Consumer | Acceptance condition |
| Updated repository documentation | Maintainers and agents | Counts, paths, terminology, commands, and release rules match the implemented state. |
Evidence Produced
| Evidence | Consumer | Acceptance condition |
| Documentation consistency check | Release owner | Affected files and verified facts are listed with no unresolved contradiction. |
Capability Contract
Read and search are required. Editing is permitted only when documentation updates are authorised by the parent task. Publishing, external messaging, deletion, and release claims require explicit authority and validation evidence.
Degraded Mode
Fallback: when a source document, diff, command, or machine result is unavailable, update only verified facts, mark the remainder not assessed, and do not claim repository-wide consistency.
Decision Rules
| Change | Action | Failure or risk avoided |
| Routing or active-count change | Update agent routers and catalogue documentation | Agents loading stale paths |
| Validator or release-gate change | Update authoring, contribution, and CI instructions | Unreproducible releases |
| Copy-only correction with no behavioural effect | Edit the owning document only | Unnecessary documentation churn |
References
- Repository authoring standard
- Contribution and release procedure
<!-- dual-compat-end -->Update project documentation systematically after significant changes. Keep all files consistent and accurate.
Core Principle: Documentation tells one cohesive story. Each file serves a specific audience but must reflect the same reality.
Modularize Instructions (Token Economy): Avoid packing everything into a single CLAUDE.md. Prefer multiple focused docs (e.g., docs/setup.md, docs/api.md, docs/workflows.md) and reference them only when needed to reduce context bloat.
When to Use
✅ Adding/removing features ✅ Architecture or design pattern changes ✅ Dependency or tech stack updates ✅ API endpoints or database schema changes ✅ Project directory restructuring ✅ Development workflow changes
❌ Typo fixes (do directly) ❌ Code comments ❌ WIP features not yet merged
Documentation Files
| File | Audience | Purpose |
| PROJECT_BRIEF.md | Stakeholders, new devs | 30-sec overview |
| README.md | Developers | Setup, usage guide |
| TECH_STACK.md | Developers, DevOps | Tech inventory |
| ARCHITECTURE.md | Senior devs, architects | System design |
| docs/API.md | API consumers | API reference |
| docs/DATABASE.md | Backend devs, DBAs | Schema docs |
| CLAUDE.md | Claude Code | Dev patterns |
| docs/setup.md | Developers | Setup details |
| docs/api.md | API consumers | API usage guide |
| docs/workflows.md | Developers | Workflow rules |
Change → File Mapping
New Feature:
- README.md (usage)
- docs/API.md (if adds endpoints)
- docs/DATABASE.md (if adds tables)
- ARCHITECTURE.md (if adds components)
- CLAUDE.md (if changes patterns)
- PROJECT_BRIEF.md (if significant)
Tech Stack Change:
- TECH_STACK.md (always)
- README.md (setup instructions)
- ARCHITECTURE.md (if affects design)
- CLAUDE.md (if affects workflows)
Architecture Change:
- ARCHITECTURE.md (always)
- README.md (overview section)
- CLAUDE.md (patterns)
- PROJECT_BRIEF.md (if major)
API/Database Change:
- docs/API.md or docs/DATABASE.md (always)
- ARCHITECTURE.md (if changes contracts)
- CLAUDE.md (if affects patterns)
- README.md (if affects usage)
Update Workflow
1. Understand Change (2-5 min)
Document:
- Type: Feature/Architecture/Tech Stack/API/Database
- What: One sentence description
- Impact: Who/what affected
- Breaking: Yes/No (what breaks)
2. Map to Files (1-2 min)
Order: Specific → General
- Technical Specs (API.md, DATABASE.md)
- Architecture (ARCHITECTURE.md, TECH_STACK.md)
- AI Instructions (CLAUDE.md)
- User Guides (README.md)
- Overview (PROJECT_BRIEF.md)
3. Read Current State (2-3 min)
Read all affected files in parallel.
4. Update Systematically (10-20 min)
Per-file checklist:
- Update primary section
- Update related sections
- Update examples/code snippets
- Add migration notes if breaking
5. Verify Consistency (2-3 min)
Check across all files:
- Terminology consistent
- Version numbers match
- File paths consistent
- Component names consistent
- Features described consistently
6. Final Review (1 min)
- New dev can understand from README
- CLAUDE.md has context
- Breaking changes marked
- Examples work
- No contradictions
Total: 15-30 minutes
Common Mistakes
❌ Updating only one file
# Updated README but forgot CLAUDE.md
# Result: AI doesn't know new pattern❌ Inconsistent terminology
# README.md: "Authentication Service"
# ARCHITECTURE.md: "Auth Module"
# CLAUDE.md: "Login System"
# Pick ONE term everywhere❌ Forgetting breaking changes
# Renamed API endpoint but README examples still use old path
# Add migration notes EVERYWHERE affected❌ General → Specific order
# BAD: Update BRIEF first, then API.md
# GOOD: Update API.md first (precise), then BRIEF (summary)Quick Reference
Update Order:
API/DB Specs → Architecture → CLAUDE → README → BRIEFConsistency Checks:
Terminology, Versions, Paths, Names, FeaturesTime Budget:
Small change: 5-10 min
Medium change: 15-30 min
Major refactor: 45-60 minSummary
Process: Understand → Map → Read → Update → Verify → Review
Key Rules:
- Update specific docs first, general last
- Read all affected files before editing
- Keep terminology consistent
- Mark breaking changes everywhere
- Test examples before committing
- One reality, multiple perspectives
Remember: Documentation debt compounds fast. Update immediately when making changes.
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.

