# Kill Review

> Test a venture against its own kill criteria, and stop it if they are met.

- Skill: `tinyhumansai/kill-review` (Agent Skill)
- Install (CLI): `npx skillmds@latest add tinyhumansai/kill-review`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tinyhumansai/kill-review/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: tinyhumansai (https://skillmd.com/u/tinyhumansai)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/tinyhumansai/kill-review

---


# Kill Review

Every venture feels alive from the inside, because there is always one more
experiment. This is the pass that asks whether the criteria the studio agreed
in advance have been met.

## When to use

- A quarter turns.
- A validation failed.
- A venture has been "nearly there" for two reviews running.

## Steps

1. **Read `kill_criteria` as written,** before looking at anything else.
   Reading them after the current numbers is how they get reinterpreted.
2. **Check them literally.** If they are met, the venture stops. That is the
   whole function of having agreed them.
3. **Check capacity honestly.** What the venture is consuming, and what the
   other ventures would do with it. This is the real question and it is rarely
   asked directly.
4. **Look for the extension pattern.** "One more experiment" twice in a row is
   itself the finding.
5. **If it continues, restate what would end it** and by when. A venture
   continuing without a refreshed criterion has quietly lost the criterion.
6. **If it stops, close the row with the reason,** and register anything reusable
   on `shared-assets`. A killed venture usually leaves something worth keeping.

## Output

A recommendation to continue with restated criteria, or to stop with the
criterion that was met — plus any shared assets salvaged. The capital decision
is the operator's; the honest read is the studio's.