From blueprint
Audit implementation against the plan — compare planned vs actual changes, detect orphaned test references, fix checkbox mismatches, and sync frontmatter counts. Use this skill whenever the user says "/plan-check", "check the plan", "audit the plan", "verify implementation", "compare plan vs code", or any request to validate that what was built matches what was planned. Also triggers on "orphaned tests", "plan audit", "check task marks", "did we implement everything", or "sync frontmatter counts". ALWAYS run after /plan-approved and before /pr — this is the quality gate.
npx claudepluginhub skaisser/blueprint-pluginThis skill uses the workspace's default tool permissions.
Read `blueprint/.config.yml` → `language`. If `auto`, detect from the user's messages. All generated content MUST be in the detected language. Skill instructions stay in English — only output changes.
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
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.
Analyzes competition with Porter's Five Forces, Blue Ocean Strategy, and positioning maps to identify differentiation opportunities and market positioning for startups and pitches.
Share bugs, ideas, or general feedback.
Read blueprint/.config.yml → language. If auto, detect from the user's messages. All generated content MUST be in the detected language. Skill instructions stay in English — only output changes.
Analyze and update the plan file for the current branch, comparing planned vs actual implementation.
/plan → /plan-review → /plan-approved → /plan-check → /pr → /review → /address-pr → /finish
[ ]. The user decides whether to implement or descope.[x] task as [ ], verify via git diff that the planned file was truly NOT modified. If a file was modified but differently than planned, keep [x] and note the deviation — do not revert to [ ].Run: echo "🔷 BP: plan-check [1/3] reading plan context + diffs"
Run these simultaneously in a single step (do not wait between them):
~/.blueprint/bin/blueprint meta — returns branch, base_branch, plan_file, project as JSON~/.blueprint/bin/blueprint context --diffs — returns commits, changed files, diff stat, AND per-file diffsIf $ARGUMENTS is a plan path or number, locate that file directly instead of using plan_file from meta.
READ the full plan file (after meta returns the path) — DO NOT proceed without reading it.
[x]/[ ] marks accurate?Agents sometimes delete [ ] tasks they couldn't solve instead of reporting failure. Compare the plan at plan-review time vs now to catch any removed tasks.
PLAN_REVIEW_COMMIT=$(git log --oneline | grep "plan: review" | head -1 | awk '{print $1}')
# PLAN_FILE already available from Step 1's blueprint meta output
echo "Comparing plan at $PLAN_REVIEW_COMMIT vs current"
# Tasks at plan-review time (the approved baseline)
git show "$PLAN_REVIEW_COMMIT:$PLAN_FILE" 2>/dev/null | grep -E "^- \[[ x]\]" > /tmp/plan-review-tasks.txt
# Tasks now (after execution)
grep -E "^- \[[ x]\]" "$PLAN_FILE" > /tmp/plan-current-tasks.txt
diff /tmp/plan-review-tasks.txt /tmp/plan-current-tasks.txt
Look for lines only in the review version (prefixed with <) — these are tasks that were deleted during execution.
If any tasks were removed:
[ ] with a note: (restored by plan-check — removed during execution)This is a hard failure — deleted tasks indicate an agent tried to hide incomplete work.
This step catches test files that existed at plan-review time, referenced removed behaviors, but were NOT modified during implementation.
PLAN_REVIEW_COMMIT=$(git log --oneline | grep "plan: review" | head -1 | awk '{print $1}')
echo "Plan-review baseline: $PLAN_REVIEW_COMMIT"
From the diffs gathered in Step 1, identify significant removed patterns (lines starting with -):
name, id, type, data, test, user). Use specific identifiers only (e.g., old_field_name, legacyEndpoint, validateOldRule).Use specific, non-generic patterns to avoid false positives.
git grep -l "REMOVED_PATTERN" $PLAN_REVIEW_COMMIT -- "tests/"
Flag any test file that contained the removed pattern at plan-review time AND was NOT modified during implementation. These are orphaned test references.
Fix all orphaned references before proceeding to the audit commit.
[x]/[ ] mismatchesdate "+%d/%m/%Y %H:%M"session field set by /plan-approvedIf the plan has an ## Acceptance Criteria section, verify each item against the actual implementation:
[x] with timestamp: - [x] Criterion ✅ DD/MM/YYYY HH:MM[ ] if they fail — note what failedRun: echo "🔷 BP: plan-check [2/3] audit commit — full task list with all marks visible"
Count mismatches fixed, deleted tasks restored, and orphaned refs fixed during this audit to build a descriptive commit message:
# If orphaned test fixes or other code changes were made, run /ship first
# Then always commit the updated plan to the project repo:
git add blueprint/ && git commit -m "🧹 chore: plan check NNNN — fixed N mismatches, restored N deleted tasks, fixed N orphaned refs"
The plan file MUST have the full task list with all [x] marks and timestamps before pushing.
Run: echo "🔷 BP: plan-check [3/3] plan check complete"
Plan Check Complete:
- Planned items: X/Y implemented
- Deleted tasks: N (restored — agents removed instead of implementing)
- Additional items: N (beyond plan)
- Missing items: N (not implemented)
- Acceptance Criteria: X/Y verified
- Files planned: X | Files modified: Y
- Status: [All matched / Discrepancies found]
STOP. You MUST use AskUserQuestion tool here.
Use $ARGUMENTS as plan file path if provided, otherwise auto-detect from branch.