agentleFS
Sign inStart free
IgorGanapolskyIgorGanapolsky/ThumbGate/.agents/skills/typesafe-typed-questions-not-clone/SKILL.md

typesafe-typed-questions-not-clone

Skill29 starsChanged 7 months agoMITView on GitHub

TypeSafe (console.typesafe.ai/hook, Jev System One) is a calibrated decision model, not a ThumbGate clone. Steal typed noul/choice/score + code-owned route() + confidence-as-second-axis onto existing PreToolUse rails. Never install typesafe-sdk, call api.typesafe.ai, clone Jev, or unpark the LLM adjudicator (#3690/#3687). Slash: /typesafe-typed-questions-not-clone.

  • Deletes or force-pushes
  • Installs packages
  • Commits and pushes

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do NOT install typesafe-sdk.

    line 33
  2. Very specific:

    Put the PreToolUse payload in state.

    line 71
  3. Distinctive:

    Ask independent typed questions (noul per hazard, choice for family, score for severity).

    line 72

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

TypeSafe typed questions — compare, do not clone

Goal

Produce fail-closed typed-question composition for whom: ThumbGate PreToolUse — so a tool-call state is scored with independent noul/choice/score questions and code routes pass|review|block. TypeSafe/Jev stays a neighbor FORMAT, not a SKU, not an LLM adjudicator.

Constraints

NEVERALWAYS
Install typesafe-sdk / wire Jev as the PreToolUse gatenpx thumbgate typesafe-typed-questions --json
--use-typesafe-api (gate)--live shadow vs POST /v1/systemone; code still owns route()
Clone Jev / System One as the gateDeterministic matchers answer the battery
Unpark LLM adjudicator (#3690/#3687)Code owns route(); model does not emit the verdict
One free-form "should we allow this?" judgeAtomic noul per hazard + score for severity
Dual-edit untracked llm-adjudicator theaterMap onto existing gate-check rails
N HTTP calls (one question each)One POST with the whole battery (parallel_questions)

HARD fail closed. REFUSE SKU clones. ECI: no net-new governance product. Do NOT install typesafe-sdk. Do NOT let Jev emit permissionDecision. --live is a shadow scorer (grok-fleet key), not the gate.

Complementary to trading /typesafe-system-one-not-clone (AGENT-651 claim gate). Do not dual-edit IgorGanapolsky/trading scripts/typesafe_claim_gate.py. This skill is ThumbGate PreToolUse only.

Reference

Examples (show, don't tell)

Weak: Summarize Jev and npm install typesafe-sdk as the new PreToolUse hook.

Gold:

$ npx thumbgate typesafe-typed-questions --json --map-only
codeOwnsRoute: true
$ npx thumbgate typesafe-typed-questions --json --tool-name=Bash --command='git push --force origin main'
{"route":"block","answers":{"destructive":{"noul":1,"source":"deterministic"}}}
$ node --test tests/typesafe-typed-questions.test.js
# --clone-jev / --use-typesafe-api / --llm-adjudicate → fail

Procedures

npx thumbgate typesafe-typed-questions --json --map-only
npx thumbgate typesafe-typed-questions --json --tool-name=Bash --command='git push --force origin main'
npx thumbgate typesafe-typed-questions --json --clone-jev
npm run test:typesafe-typed-questions
  1. Put the PreToolUse payload in state.
  2. Ask independent typed questions (noul per hazard, choice for family, score for severity).
  3. Answer them with deterministic matchers — never Jev.
  4. route() in code: action threshold → block, review threshold → review, severity ≥ 2 promotes review to block.
  5. Refuse --clone-jev, --use-typesafe-api (gate), --llm-adjudicate, and model-emitted verdicts.
  6. Optional --live shadows the same battery against Jev in one POST. --fan-out-questions fails. Receipt is estimated N× input tokens, not the cookbook's 12.2×.

Rubric

  • empty / ordinary Read → route=pass, ok=true
  • git push --force → route=block, destructive.noul=1
  • git add -A → route=review (warn-level noul=0.55)
  • --clone-jev / --use-typesafe-api / --llm-adjudicate → ok=false
  • playground-shaped noul without a matcher → unevaluated_question (do not call Jev)
  • doctor: npm run test:typesafe-typed-questions PASS
  • evidence: command output in the same turn

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free