Orchestrate documentation review by assigning docs-curator agents to individual documents for thorough validation against the codebase.
From project-docsnpx claudepluginhub ichabodcole/project-docs-scaffold-template --plugin project-docsThis skill uses the workspace's default tool permissions.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
Retrieves current documentation, API references, and code examples for libraries, frameworks, SDKs, CLIs, and services via Context7 CLI. Ideal for API syntax, configs, migrations, and setup queries.
Uses ctx7 CLI to fetch current library docs, manage AI coding skills (install/search/generate), and configure Context7 MCP for AI editors.
A management skill for systematically reviewing documentation. Assign one docs-curator agent per document to ensure thorough investigation.
Activate when:
Key indicators:
Assigning multiple documents to a single agent dilutes review quality. Each document requires:
Always assign one docs-curator agent per document.
List the documents that need review:
# List project proposals
ls docs/projects/*/proposal.md
# List project plans
ls docs/projects/*/plan.md
# List architecture docs
ls docs/architecture/*.md
# List specifications
ls docs/specifications/*.md
Or use Glob to find specific patterns:
docs/projects/*/proposal.md - All project proposalsdocs/projects/*/plan.md - All project plansdocs/projects/*/sessions/*.md - All project sessionsdocs/architecture/*.md - All architecture docsdocs/specifications/*.md - All specificationsDifferent document types have different review goals:
Temporal Documents (check for completion/archival):
docs/projects/*/proposal.md - Archive project when implementeddocs/projects/*/plan.md - Archive project when completedocs/investigations/ - Archive when concludeddocs/reports/ - Archive when acted uponEvergreen Documents (check for accuracy):
docs/architecture/ - Update to match current codedocs/specifications/ - Update to match current application behaviordocs/playbooks/ - Update if procedures changeddocs/lessons-learned/ - Update if solutions evolveddocs/interaction-design/ - Update to match current UXLaunch one docs-curator agent per document using the Task tool:
Task(
subagent_type: "docs-curator",
description: "Review [document-name]",
prompt: "Review this document for accuracy and implementation status:
**Document:** docs/projects/feature-x/proposal.md
Read the document, search the codebase for evidence of implementation,
check git history for related commits, and provide your findings with
the recommended action (archive, update, no action)."
)
For parallel efficiency, launch multiple agents in a single message:
// Launch 4 agents in parallel, one per document
Task(...doc1...)
Task(...doc2...)
Task(...doc3...)
Task(...doc4...)
As agents complete, collect their findings:
Group by recommended action:
Present summary to user:
## Documentation Review Results
### Ready to Archive (5 documents)
- projects/feature-x/ - Fully implemented
- projects/feature-y/ - Superseded by feature-z ...
### Needs Update (2 documents)
- architecture/api-design.md - Section 3 outdated ...
### Current (3 documents)
- playbooks/deployment.md - Accurate ...
Get user approval before taking action
For approved archives:
_archive/ subfolderFor approved updates:
For large documentation sets:
| Doc Count | Approach |
|---|---|
| 1-5 | Review all in parallel |
| 6-15 | Batch into 2-3 waves |
| 16+ | Prioritize by age/importance, review subset |
Prioritization criteria:
Copy and customize for each document:
Review this document for accuracy and implementation status:
**Document:** [full path to document]
**Document Type:** [Proposal | Plan | Architecture | etc.]
Tasks:
1. Read the full document
2. Search codebase for mentioned features/components
3. Check git log for related commits
4. Determine implementation/accuracy status
5. Recommend action: Archive | Update | No Action
Provide findings in the standard docs-curator output format.
# 1. List active project proposals (excluding archive)
ls docs/projects/*/proposal.md | grep -v archive
# Result: 8 projects to review
# 2. Launch 4 agents in parallel (first batch)
Task(subagent_type: "docs-curator", description: "Review project-a proposal", prompt: "...")
Task(subagent_type: "docs-curator", description: "Review project-b proposal", prompt: "...")
Task(subagent_type: "docs-curator", description: "Review project-c proposal", prompt: "...")
Task(subagent_type: "docs-curator", description: "Review project-d proposal", prompt: "...")
# 3. After first batch completes, launch remaining 4
Task(subagent_type: "docs-curator", description: "Review project-e proposal", prompt: "...")
...
# 4. Consolidate results and present to user
"8 projects reviewed:
- 5 ready for archive (fully implemented)
- 2 partially complete (list what remains)
- 1 not started
Would you like me to archive the 5 completed project folders?"
After consolidating results, generate a report document:
docs/reports/YYYY-MM-DD-doc-status-report.mdReport sections: