Notre avis
Construit un fichier context.md pour une tâche en lisant la documentation structurée du module et les fichiers de code pertinents.
Points forts
- Automatise la collecte de contexte à partir de multiples sources (README, fonctionnalités, code).
- Gère les dépendances inter-modules et les contraintes de fonctionnalités existantes.
- Réduit la charge cognitive en présentant uniquement les informations pertinentes.
Limites
- Nécessite une structure de projet bien définie (workflow.yaml) pour fonctionner.
- Peut manquer de contexte si les docs du module sont absentes ou incomplètes.
- Ne gère pas les contextes très larges ou monolithiques.
Utilisez cette compétence lorsque vous commencez une nouvelle tâche et avez besoin d'un résumé rapide du contexte du module.
Ne l'utilisez pas pour des tâches très simples où le contexte est évident, ou lorsque la documentation du module est inexistante.
Analyse de sécurité
SûrThe skill uses Bash only for read-only commands like ls, git ls-files, and git grep to gather project context. No destructive operations, external data exfiltration, or execution of arbitrary payloads. Minimal risk.
Aucun point d'attention détecté
Exemples
Run w-context-load CISTREAMIN-11-sign-up-apiw-context-load TASK-42-refactor-payment payment-modulew-context-load $(cat .claude/current-task.txt)description: Phase 0b helper for w-task — load module-aware context by reading module README + features/*.md + workflow-links, then write context.md. Generic version of dev-task's streaming-docs-aware context load. argument-hint: <task-slug> [<module-name>] allowed-tools: Bash, Read, Write, Edit
w-context-load
Build context.md for the current task by reading structured module docs
(README, features/*.md, workflow-links.md) when they exist, falling back to
git-grep file inventory when they don't.
Reads .claude/workflow.yaml to find:
workflow.module_docs_rootworkflow.feature_records_subdirworkflow.workflow_links_filenameproject.module_globproject.src_dirs
Inputs
$1— task slug (e.g.CISTREAMIN-11-sign-up-api)$2(optional) — module name override; otherwise inferred from intake.md
Reads <state_root>/<task-slug>/intake.md for title + description.
Step 1 — Resolve module
If $2 is provided, use it.
Otherwise, infer from intake.md:
- Extract 3-5 keywords from title + description (nouns).
- If
module_docs_rootset:ls <module_docs_root>/and match keywords against folder names. - Fallback: read first 30 lines of intake.md, look for explicit "Module: X" line.
- If still ambiguous: pause, ask developer to pick from listed modules.
Write state.yaml.module: <name> once resolved.
Step 2 — Read module README (if exists)
If <module_docs_root>/<NN>-<module>/README.md exists:
- Parse "Implementation Status" section — extract ✅ Complete and ⏳ Deferred lists.
- Parse "Depends on" / "Blocks" — cross-module deps.
- Note last update date and any "Last modified by task: <slug>" markers.
If absent: note "Module README not found — context will be code-only."
Step 3 — Read feature records
If <module_docs_root>/<NN>-<module>/<feature_records_subdir>/ exists:
lsthe subfolder, list all*.mdfiles.- For each file, read the "Constraints for future features" section.
- Cross-reference constraints against the current task keywords.
- Collect any "Conflict candidates" — features that touch overlapping concerns.
Step 4 — Read code (filtered by README)
Use the Implementation Status ✅ Complete list to scope file reads:
- For each ✅ item, find the matching file under
project.module_glob/project.src_dirsand read it. - Skip ⏳ Deferred items — they don't exist yet.
If no README available, fall back to generic Phase 0b behaviour:
git ls-files | head -300
git grep -l "<keyword>" 2>/dev/null | head -20
Read top 3-5 matched files.
Step 5 — Reuse check
Scan for cross-cutting utilities under project.src_dirs (e.g. src/common/,
src/shared/):
- Common guards, decorators, exceptions
- Shared services
- Already-installed libs from
package.json/ equivalent
Step 6 — OQ surfacing
If workflow.oq_docs_path is set: invoke w-oq-check <module-or-task-slug>
and append its output under § Known Constraints in context.md.
Step 7 — Write context.md
# Context — <title>
## Task Summary
<intake title, ticket ID, module, developer notes>
## Module Snapshot
**Implementation Status (from README):**
- ✅ <list>
- ⏳ <list>
**Depends on:** <list>
**Blocks:** <list>
## Relevant Files
| File | Why relevant |
|------|-------------|
| ... | ... |
## Patterns Observed
<naming conventions, existing similar features>
## Available Utilities
<reusable guards/decorators/libs found in Step 5>
## Previous Feature Decisions
<constraints from features/*.md — omit section if none>
## Known Constraints
<OQs from w-oq-check + flags from README>
## Suggested Approach
<1-2 sentences>
Step 8 — Update workflow-links
If workflow.module_docs_root + workflow_links_filename set:
- Path:
<module_docs_root>/<NN>-<module>/<workflow_links_filename> - Create from template if absent:
# Workflow Links: <module> | Task | Folder | Started | Status | |------|--------|---------|--------| - Append row:
| <task-slug> | [<state_root>/<task-slug>/](...) | <date> | in-progress |
Output
Print to chat: "Context loaded for module <name>. Found N relevant files,
M previous feature decisions, K known constraints. Run /w-task to start planning."
Expert Next.js App Router
Developpement
Un skill qui transforme Claude en expert Next.js App Router.
Générateur de README
Developpement
Crée des README.md professionnels et complets pour vos projets.
Rédacteur de Documentation API
Developpement
Génère de la documentation API complète au format OpenAPI/Swagger.