Is Arm Evals safe and trustworthy?

tashan scores Arm Evals on public evidence — see the measured signals. Audit scope: OSV advisories, install-time scripts, build provenance and declared permission surface. We do not review source code or execute the capability.

How do I install Arm Evals?

Drop the skill folder into ~/.claude/skills/ (user scope) or .claude/skills/ (project scope).

Who maintains Arm Evals?

source: Aznatkoiny/claude-dev-toolkit.

‹ The Index

Arm Evals

skill

Arm a forge eval suite into a frozen build contract. Run when .forge/state.json shows phase=evals, after /plugin-forge:build-evals; re-run after any suite edit or triage grader-error to re-version and re-freeze. Four gates in order via forge-eval: reference (every task's reference solution passes its own graders), red (no-plugin baseline; tasks that pass bare carry no signal), metaeval (judge vs human agreement = 0.85 against evals/labels/labels.jsonl), then sha256 FREEZE into .forge/freeze.json with suiteversion vN and phase - armed. Refuses to freeze uncalibrated judges, unresolved no-signal tasks, or a suite with pending change requests.

Works with: Claude Code (native) · Cursor, Codex CLI (manual)
native: this artifact type is that client's own format

Category: Communication — see all ranked ›

Install (Claude Code):

cp -r arm-evals ~/.claude/skills/

Adoption: 1 repos

Security audit

Not scanned yet. We audit npm-published capabilities for known advisories, install-time scripts and permission surface; this one has no npm package we can resolve, or has not reached the queue.

source ↗ · skill:Aznatkoiny/arm-evals

Already running this? npx tashan-cli doctor checks your whole config against the Index — how it works ›

Measured 2026-08-05 · scorer s5 · how · something wrong here?