plan-founder-review — Claude Code skill: audit a plan before agents run
plan-founder-review reads a plan and its JSON sidecar, checks every path and reuse claim against the real repository, then returns APPROVE, REVISE or REJECT before any agent starts work.
no license 170 installs ★ 5 updated 2026-08-31
npx skills add ulpi-io/skills --skill plan-founder-review -a claude-codeWhat plan-founder-review does
Point it at a plan in .ulpi/plans/ or give a direct path. The skill opens the whole markdown file and the matching JSON, then walks the repo to confirm that files listed for editing are really there and that files to be created have a parent that exists or a task that makes them first. Claims that some module can be reused get the same treatment: the code either exists or it doesn't.
Two modes. --quick stops after architecture and the dependency graph; --full also covers failure handling, rollback, test coverage and whether the task breakdown is realistic for agents. Sections are scored PASS, WARN or FAIL, findings are split into blocking and minor, and the summary is a single verdict. The detailed rules live in references/review-checklist.md and references/review-rubric.md.
It runs in a forked context with its own reasoning budget, and the tool set is read-only: Read, Glob, Grep, Bash. Your plan file comes back untouched, so fixes after a REVISE are yours to make. No plan at the target path means the review stops right away.
What it helps with
- catch file paths in a plan that don't exist in the repo
- spot drift between the plan markdown and its JSON
- verify that code the plan wants to reuse is actually there
- judge whether risk notes and test coverage are good enough before agents start
- get a verdict on a plan without running it
Example prompts for the agent
- Review the
auth-systemplan before we hand it to the agents - Check
.ulpi/plans/foo.mdin--fullmode - Run a quick review of the latest plan
How to install plan-founder-review
1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill plan-founder-review -a claude-code2Manually: put the files in place
~/.claude/skills/plan-founder-review/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/plan-founder-review/
1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill plan-founder-review -a codex2Manually: put the files in place
.agents/skills/plan-founder-review/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill plan-founder-review -a cursor2Manually: put the files in place
.cursor/skills/plan-founder-review/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install ulpi-io/skills plan-founder-review2Manually: put the files in place
.github/skills/plan-founder-review/SKILL.md1Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill plan-founder-review -a gemini-cli2Platform CLI
gemini skills install https://github.com/ulpi-io/skills --path plan-founder-reviewManually: put the files in place
.devin/skills/plan-founder-review/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add ulpi-io/skills --skill plan-founder-review -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about plan-founder-review
How do I install plan-founder-review in Claude Code?
Run npx skills add ulpi-io/skills --skill plan-founder-review -a claude-code. The skill lands at ~/.claude/skills/plan-founder-review/SKILL.md.
What does plan-founder-review do?
It takes a plan plus its JSON, verifies paths and reuse claims against the actual repo, scores the sections and returns APPROVE, REVISE or REJECT. The plan file itself stays as it was.
How is this different from a code review?
It audits the plan artifact before execution rather than a diff. Plans are generated by plan-to-task-list-with-dag, and bugs in code changes are find-bugs territory.