From devproc
Sync all documentation and feature tracking to reflect the current state of the in-progress feature
How this skill is triggered — by the user, by Claude, or both
Slash command
/devproc:feature-checkpointThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Run this skill after every sub-task completes — do not wait to be asked.**
Run this skill after every sub-task completes — do not wait to be asked.
Bring all project documentation up to date with the current state of the feature.
Before proceeding, check that CLAUDE.md contains a Feature Model section. If it doesn't, tell the user to run /feature-init first and stop.
The feature to checkpoint is: $ARGUMENTS (if blank, use whichever feature is currently In Progress in FEATURES.md)
Steps:
Read FEATURES.md to identify the in-progress feature and its plan file reference.
Read the plan file (plans/<slug>.md). Update it to reflect the actual current state:
3. **Refactor auth module** — extract session handling to its own class
- ✓ Extracted SessionManager class
- ✓ Moved token refresh logic
- ▶ TODO: Update tests to use new interface
Update the ## Handoff section at the top of the plan file. Create it if it does not exist. This section is the single source of truth for resuming work in a new session:
## Handoff
**Last updated:** YYYY-MM-DD
**Session summary:** [1–2 sentences on what this session accomplished]
**Sub-task in progress:** Sub-task N — [what has been done within it so far, if partially complete]
**First action next session:** [specific, concrete first step — not "continue sub-task N" but the actual action]
**Open questions / decisions pending:** [anything unresolved that needs a decision]
**Dead ends to avoid:** [approaches tried and abandoned this session, with reasons — omit if none]
Be specific and concrete. Vague entries like "continue working on the feature" are useless to a cold session.
Check the documentation for all user and architectural documentation in the various docs directories; they should be kept up to date continuously.
Fix any drift.
Check NOTES.md. For each non-obvious technical finding made since the last checkpoint (page structure quirks, API behaviour, design decisions, scope changes), add a note.
Check CLAUDE.md current status section. Update it to accurately reflect what is done and what is next. Keep it high-level — one paragraph per component.
Verify FEATURES.md In Progress entry is still accurate (description has not drifted from what the feature is actually doing).
Report a brief summary of what was updated.
Rule: do not wait to be asked. This skill exists because documentation and tracking information drifts silently. Run it proactively when a sub-task completes, and whenever requested.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
npx claudepluginhub plwhite/claudeplugins --plugin devproc