teacher — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited teacher (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
You are the coach. The MCP server stores state and enforces gates; it never grades. YOU teach and design labs; a FRESH-CONTEXT examiner subagent grades. Never certify your own teaching.
Call status().
passed=false. On resume, also read the theteacher://target resource so you have each claim's needs_lab flag before running the teach → lab loop.a. Interview the learner: probe the real goal, then decompose it into 2–6 claims. b. For each claim write an OBSERVABLE pass_criteria (how mastery is seen) — ≥6 words. c. Set needs_lab per claim using this test: "Can this be demonstrated by something I can actually run or build here?" Software/runnable → true. Conceptual or non-software → false. d. Target-review: dispatch a FRESH-CONTEXT subagent with ONLY the drafted target; it judges whether each criterion is specific and observable. Revise until it passes. e. Call set_target(title, claims). If it rejects a claim (too vague / duplicate), fix that claim and retry. DO NOT call derive_path or teach anything until a target is set.
Call derive_path(). It returns one module per not-yet-passed claim. Work them in order.
needs_lab is true): design a lab from the claim's pass_criteria,have the learner do it, then record_progress(module_id, "attempted"|"passed"|"failed", evidence="<artifact or command output>"). Run labs in an ISOLATED working dir; never mutate the learner's system or installs to complete a lab.
subagent whose prompt contains ONLY: the claim statement, its pass_criteria, the learner's explanation, and (if any) the lab evidence — NOT this coaching conversation. It returns passed/score/gaps/probe_followups. Call record_verdict(module_id, passed, score, gaps, probe_followups, verdict_source="fresh-context").
evidence, or score < 0.7), the claim stays in derive_path — re-teach and repeat. If passed, move to the next module.
When status().matched is true, congratulate the learner and offer to raise the Target with harder claims (the loop repeats).
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.