Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →/cs:freeze <decision> <days> — Lock a strategic decision for a cooldown period to prevent impulse reversal. Mirrors gstack's safety primitives for the business layer. Use when an irreversible decision was made under pressure — e.g. a layoff plan or multi-year contract — and deserves a cooling-off lock before execution.
.claude/skills/alirezarezvani-freeze/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 56% | 0% |
Command: /cs:freeze <decision-path> <days>
Locks a decision for a defined cooldown period. During the freeze, the chief-of-staff router refuses to re-litigate the decision unless a kill criterion explicitly triggers.
Inspired by gstack's /freeze and /guard safety primitives — adapted from code-scoping to strategic-scoping.
Founders are pattern-matchers; pattern-matching after a tough decision often produces a reversal that's actually just decision fatigue. The freeze enforces a discipline:
| Decision type | Default freeze | |---|---| | Fundraise round size / lead choice | 30 days | | Pricing change | 60 days | | Market entry / exit | 90 days | | Layoff / RIF | 30 days | | Strategic pivot | 90 days | | Personnel (exec hire / fire) | 60 days | | M&A LOI | 30 days | | Custom | specify in command |
freeze_until: YYYY-MM-DD to the decision record~/.claude/freezes/active.mdThe decision record is updated in place:
markdown# Decision: <title> ... **Status:** FROZEN **Frozen until:** YYYY-MM-DD **Reason for freeze:** <text> **Override condition:** Kill criterion <name> triggers OR founder issues `/cs:unfreeze` with stated reason
The active-freezes index is updated:
markdown# Active Freezes **Updated:** YYYY-MM-DD | Decision | Frozen until | Override condition | |---|---|---| | <decision title> | YYYY-MM-DD | <kill criterion or /cs:unfreeze> |
To unfreeze before the period ends, the founder runs:
/cs:unfreeze <decision> <reason>The unfreeze is logged in the decision history (preserved permanently). Forced overrides create a paper trail that surfaces at post-mortem.
If a kill criterion in the decision triggers, the freeze auto-releases and the chief-of-staff routes immediately to /cs:post-mortem. The freeze does not protect against reality; it protects against impulse.
Founders have authority. Without an explicit lock + log, every wobble produces a "let's discuss this again" — which is exhausting for advisors and erodes the value of the boardroom. The freeze is a process, not a rule; it logs every override so the post-mortem can audit founder discipline.
/cs:unfreeze — explicit early release/cs:post-mortem — auto-triggered if kill criterion fires/cs:boardroom — blocked until unfreeze or expirydecision-loggercs-chief-of-staff — enforces freezes in routingVersion: 1.0.0
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→pass | 8,462 | 12,198 | +44% | 1 | 1 | 0% | 1,445 | 2,307 | +60% | 0 | 0 | — |
case-03 | fail→fail | 17,888 | 9,631 | -46% | 1 | 1 | 0% | 2,150 | 1,896 | -12% | 0 | 0 | — |
case-12 | fail→pass | 12,131 | 11,355 | -6% | 1 | 1 | 0% | 1,943 | 1,997 | +3% | 0 | 0 | — |
case-01 | fail→pass | 15,387 | 8,030 | -48% | 1 | 1 | 0% | 3,097 | 2,333 | -25% | 0 | 0 | — |
case-04 | fail→pass | 26,315 | 8,741 | -67% | 1 | 1 | 0% | 1,890 | 1,489 | -21% | 0 | 0 | — |
case-05 | fail→pass | 10,063 | 7,143 | -29% | 1 | 1 | 0% | 781 | 1,215 | +56% | 0 | 0 | — |
case-06 | fail→pass | 17,159 | 6,938 | -60% | 1 | 1 | 0% | 2,420 | 1,264 | -48% | 0 | 0 | — |
case-07 | fail→pass | 14,475 | 8,618 | -40% | 1 | 1 | 0% | 1,187 | 1,550 | +31% | 0 | 0 | — |
case-08 | fail→pass | 18,398 | 8,373 | -54% | 1 | 1 | 0% | 2,142 | 1,490 | -30% | 0 | 0 | — |
case-09 | fail→pass | 12,813 | 7,863 | -39% | 1 | 1 | 0% | 1,312 | 1,423 | +8% | 0 | 0 | — |
case-10 | fail→pass | 12,130 | 4,234 | -65% | 1 | 1 | 0% | 261 | 1,436 | +450% | 0 | 0 | — |
case-11 | fail→pass | 17,279 | 3,156 | -82% | 1 | 1 | 0% | 1,908 | 1,456 | -24% | 0 | 0 | — |
case-23 | pass→pass | 20,218 | 19,862 | -2% | 1 | 1 | 0% | 2,321 | 3,357 | +45% | 0 | 0 | — |
case-13 | pass→pass | 10,130 | 8,340 | -18% | 1 | 1 | 0% | 1,714 | 2,488 | +45% | 0 | 0 | — |
case-14 | fail→pass | 10,982 | 4,338 | -60% | 1 | 1 | 0% | 1,430 | 1,696 | +19% | 0 | 0 | — |
case-15 | fail→pass | 13,786 | 2,731 | -80% | 1 | 1 | 0% | 1,237 | 1,328 | +7% | 0 | 0 | — |
case-16 | fail→pass | 21,961 | 6,822 | -69% | 1 | 1 | 0% | 2,785 | 1,114 | -60% | 0 | 0 | — |
case-17 | pass→pass | 14,619 | 9,315 | -36% | 1 | 1 | 0% | 1,404 | 1,559 | +11% | 0 | 0 | — |
case-18 | fail→pass | 16,439 | 7,436 | -55% | 1 | 1 | 0% | 1,874 | 1,328 | -29% | 0 | 0 | — |
case-19 | fail→pass | 12,010 | 7,504 | -38% | 1 | 1 | 0% | 1,304 | 1,304 | 0% | 0 | 0 | — |
case-20 | pass→pass | 12,611 | 7,488 | -41% | 1 | 1 | 0% | 1,297 | 1,260 | -3% | 0 | 0 | — |
case-21 | pass→fail | 18,289 | 16,854 | -8% | 1 | 1 | 0% | 2,919 | 3,556 | +22% | 0 | 0 | — |
case-22 | pass→pass | 11,957 | 17,982 | +50% | 1 | 1 | 0% | 2,058 | 3,080 | +50% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 21 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +65 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/10/2026 | +65% |
Other measured skills in the registry, with their headline benchmark lift.