writing-super-skills — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited writing-super-skills (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.
Creating a Super Skill IS TDD applied to process documentation.
REQUIRED BACKGROUND: Understand tdd-verified and verification-before-completion before writing skills. Same Iron Laws apply to documentation.
Authoring fundamentals: CSO (description = triggers only, never workflow summary), token efficiency (< 600 words), flowcharts only for non-obvious decisions, one excellent example over many mediocre ones. This skill focuses on Super Skills–specific requirements and VHEATM quality gates on top of those fundamentals.
A Superskills skill encodes technique. A Super Skill encodes technique + VHEATM verification gates.
Every Super Skill must have:
| Element | Superskills | Super Skills |
|---|---|---|
| Trigger condition | ✅ | ✅ |
| Workflow steps | ✅ | ✅ |
| Rationalization prevention | ✅ | ✅ |
| Evidence anchors | ❌ | REQUIRED |
| Idempotency guard | ❌ | REQUIRED |
| Loop/escalation path | ❌ | if bidirectional |
| Schema reference | ❌ | if writing to KB |
Run scenarios WITHOUT the skill. Document exactly:
Address those specific failures. Structure:
---
name: [verb-noun, e.g., audit-design]
description: Use when [triggering condition only — no workflow summary]
---
# [Skill Name]
[One-sentence core principle]
## [Trigger / When to Use]
[Condition that fires this skill, including any flag checks]
## Idempotency Check
[How to detect if this skill already ran — prevent duplicate output]
## [Main Steps]
[Numbered, each with evidence gate where output is produced]
## Output Format
[Exact format for KB artifacts — reference shared/pattern-debt-schema.md if applicable]
## Red Flags — Never
[Explicit loophole closures]Token targets: Trigger-heavy skills → < 300 words. Reference skills → < 600 words.
After GREEN, run this checklist:
Evidence gates:
Idempotency:
Loop safety (if skill escalates to another):
Schema alignment:
docs/superskills/, does it reference a canonical schema?kb-query?Rationalization table:
Skills that need deterministic triggers use spec frontmatter flags:
SPEC_APPROVED: true # triggers audit-design
SPEC_ESCALATION: true # triggers audit-design re-runIf your skill needs a similar deterministic trigger, define a flag in this style. Document it in README.md Spec Frontmatter section.
If your skill writes to docs/superskills/:
→ ADRs: docs/superskills/adrs/YYYY-MM-DD-<slug>.md
→ Pattern Debt: docs/superskills/pattern-debt.md (use canonical schema)
→ QBR calibration: docs/superskills/qbr-calibration.md
→ New file types: add to kb-query Common Query PatternsAlways commit KB artifacts before claiming step complete.
# ✅ Correct — triggering conditions only
**REQUIRED:** Use `tdd-verified` before writing implementation code
# ❌ Wrong — force-loads full skill context
@skills/tdd-verified/SKILL.mdwc -w SKILL.md)<!-- Populated by knowledge-compound skill after cycles where this skill underperformed --> <!-- Format: [YYYY-MM-DD] What failed | Root cause | What to do instead --> <!-- DO NOT pre-populate with speculation — real observations only --> <!-- Example: [2026-06-01] Agent skipped risk scoring for "trivial" CSS task that touched auth config | Triage gate not read carefully | Triage gate now checks file content, not just task description -->
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.