Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Crank · boshu2 | DecimalAI

agent toolingby boshu2

Crank

boshu2/crank

Source boshu2/agentops ↗License Apache-2.0

Execute one caller-selected wave by invoking RPI once per lane and return the wave evidence, then stop. Triggers: "crank", "execute the next wave", "run this wave".

SkillScore86Maintained

Automatic

Coding pack

Last verified 10/10/2026

What it changes

Passed 18 of 21 test cases with the skill, and 9 of 21 without it.

Without skill43%

With skill86%

Checks it passed that the model alone missed

Invokes rpi exactly once for Lane A.
Uses PASS for successful lane status.
Reports the wave as unrunnable because Lane 1 lacks a brief.

and 9 more in the full report

10 fixed8 passed either way2 still failing1 broke2 errored

Pass-rate lift: +43 pp
Tokens: +10%
Agent turns: 0%
Response time: -26%

Model gemini-3.6-flashMeasured 10/10/2026v118 of 23 cases passed with skill · 21 comparable

Tokens, turns and time describe resource use, not a dollar price.

Use this skill

Save it into your project's skills folder. Claude Code picks it up in its next session.

curl -fsSL --create-dirs -o .claude/skills/boshu2-crank/SKILL.md https://app.decimal.ai/s/boshu2-crank/SKILL.md

Want it pinned and updatable? pip install "decimalai>=0.15", then decimalai skills add boshu2-crank, and start Claude Code with claude --plugin-dir "$(git rev-parse --show-toplevel)/.agents"

1 · One-time — install the CLI (docs ↗)

pip install "decimalai>=0.15"

2 · Get the skill — writes .agents/skills/boshu2-crank/ and pins it in .decimal/skills.lock

decimalai skills add boshu2-crank

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/boshu2-crank/SKILL.md https://app.decimal.ai/s/boshu2-crank/SKILL.md

Raw file: /s/boshu2-crank/SKILL.md ↗ · machine-readable summary at /s/boshu2-crank.json ↗

Safety checks: 3 of 3 passed

Code scan CleanIntent PassedContent Clean

See it in action

One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.

1 of 10

Prompt

Process this single wave with repair_rounds=2: Lane 1 - scope `src/components/Header.tsx`, acceptance `npm test Header.test.tsx`. Wave acceptance: `npm test`. Return the resulting execution state.

What was checked, without and with the skill
Checked	Without skill	With skill
Reports the wave as unrunnable because Lane 1 lacks a brief.	Missed	Passed

case-19 · gemini-3.6-flash · 10/10/2026 · answers are capped and redacted

What's inside

The instructions your model reads when the skill loads.

SKILL.md84 linessingle file

Crank

Crank executes exactly one wave. The caller selects the wave and the repair bound; crank runs that wave and returns what happened.

caller's wave + repair_rounds -> RPI per lane -> wave acceptance once -> evidence -> stop

Input

The caller supplies all three:

The wave — a list of lanes, each with a write scope, a brief, and its own executable acceptance.
repair_rounds — the caller's repair bound, forwarded unchanged to every lane. Crank neither chooses it nor spends it.
The wave acceptance — the one command set that judges the wave as a whole.

If a lane arrives without a scope, a brief, or an executable acceptance, report the wave as unrunnable and stop. Crank does not shape the missing part itself.

Contract

Read the caller's wave as given. Do not select, extend, drop, or re-rank lanes; the wave arrives already decided.
Partition the lanes. Two lanes may run in parallel only when their write scopes are disjoint and their regen surfaces (generated outputs, mirrors, manifests) are disjoint. Any shared surface serializes them; unknown overlap counts as overlap.
Invoke rpi once per lane with that lane's intent, scope, acceptance, and the caller's repair_rounds. Each lane's traversal owns its own repair loop. Crank never repairs a lane itself, never re-invokes a returned lane, and never re-plans one.
Run the wave acceptance once, after every lane has returned. A lane that returned FAIL or NOT_PROVEN does not suppress it; both results are reported side by side.
Return the wave evidence and stop — per-lane status (PASS | FAIL | NOT_PROVEN | NOT_PLANNED | NOT_BUILT), the open findings each lane's validator left keyed by their findings[].id, each lane's

Track record

Benchmark

18 / 21

test cases passed, vs 9 / 21 without

Adoption

Users: 0
Times used: 0

Safety and source details

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v5 — 9/10/2026

View 15 scanner checks

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

boshu2 / agentops

skills-codex/crank/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 9/10/2026Version v1

More skills from the same source and nearby in the registry, with their headline benchmark lift.

Ready to supercharge your AI agents?

See it in action

What's inside

Crank

Input

Contract

Track record

Benchmark

Adoption

SkillSafety

Source

Related skills