spec-to-code-compliance
Verdict safe, but 2 findings (2 medium), mainly prompt injection, resource abuse. Most recent scan 2026-02-06.
This skill performs blockchain specification-to-code compliance analysis with a rigorous, structured approach. The primary security concerns are: (1) instruction override attempts through exhaustive behavioral rules that could bypass safety guidelines, (2) transitive trust risks from missing referenced files that could contain malicious instructions, (3) resource exhaustion from unbounded exhaustive analysis requirements, and (4) missing optional metadata. The skill does not contain data exfiltration, credential theft, or command injection threats. The referenced files that ARE present (OUTPUT_REQUIREMENTS.md, COMPLETENESS_CHECKLIST.md, IR_EXAMPLES.md) contain legitimate analysis guidance without malicious content. The main risks are prompt injection patterns and resource abuse rather than active exploitation.
Scanner's overall assessment (LLM analyzer)
Second scanner: NVIDIA SkillSpector
SkillSpector found no issues but could not fully inspect the package, usually because SKILL.md refers to a file that is not present, so it fails closed to CAUTION instead of SAFE. Read it as "not fully checked", not as a finding. How the two scanners compare across the whole corpus: Two scanners, one corpus.
What this skill does
Verifies code implements exactly what documentation specifies for blockchain audits. Use when comparing code against whitepapers, finding gaps between specs and implementation, or performing compliance checks for protocol implementations.
- Package
- 4 files, 33 KB
- SKILL.md
- 1338 words, 12+ sections
- Code
- None. Instructions only.
SKILL.md covers: When to Use · When NOT to Use · Spec-to-Code Compliance Checker Skill · GLOBAL RULES · Rationalizations (Do Not Skip) · PHASE 0 — Documentation Discovery · PHASE 1 — Universal Format Normalization · PHASE 2 — Spec Intent IR (Intermediate Representation) · PHASE 3 — Code Behavior IR · (WITH TRUE LINE-BY-LINE / BLOCK-BY-BLOCK ANALYSIS) · PHASE 4 — Alignment IR (Spec ↔ Code Comparison) · PHASE 5 — Divergence Classification
Findings
MEDIUM Potential Instruction Override via Unbounded Autonomy Prompt Injectionllm
The skill instructions contain directives that could override safety guidelines through exhaustive analysis requirements. Phrases like 'NEVER infer', 'ALWAYS cite', 'Do NOT rely on prior knowledge', and 'Be literal, pedantic, and exhaustive' combined with 'Rationalizations (Do Not Skip)' table create a framework that could be used to bypass normal reasoning constraints. The skill attempts to redefine the agent's analytical behavior with strict rules that may conflict with safety guidelines.
SKILL.md
GLOBAL RULES section contains: 'Never infer unspecified behavior', 'Always cite exact evidence', 'Always provide a confidence score', 'Do NOT rely on prior knowledge'. Rationalizations table includes entries like 'I'll note this as partial match | Partial = potential vulnerability | Investigate until full_match or mismatch' which prescribe specific behavioral overrides.MEDIUM Resource Exhaustion Risk from Exhaustive Analysis Requirements Resource Abusellm
The skill mandates exhaustive analysis with requirements like 'analyze EVERY function', 'EVERY Spec-IR item has corresponding Alignment record', 'Minimum threshold met: 10+ items for non-trivial spec', and 'Be literal, pedantic, and exhaustive'. For large codebases or complex specifications, this could lead to unbounded computation, excessive token usage, and potential timeout or resource exhaustion. The COMPLETENESS_CHECKLIST.md contains 30+ mandatory checkboxes that must all be verified.
SKILL.md
GLOBAL RULES state 'Be literal, pedantic, and exhaustive'. OUTPUT_REQUIREMENTS.md mandates 'MUST analyze EVERY function', 'EVERY Spec-IR item has corresponding Alignment record', 'Minimum threshold met: 10+ items'. COMPLETENESS_CHECKLIST.md contains extensive verification requirements with no bounds on analysis scope.What is in the package
| File | Kind | Size |
|---|---|---|
SKILL.md | doc | 10 KB |
resources/COMPLETENESS_CHECKLIST.md | doc | 3.2 KB |
resources/IR_EXAMPLES.md | doc | 15 KB |
resources/OUTPUT_REQUIREMENTS.md | doc | 5.3 KB |
Package contents recorded 2026-10-22 at commit a50696e20581. The scan above ran against the version current at scan time; the two can differ if the skill has since been updated.
What we checked
- Static analyzer. Pattern rules over SKILL.md and bundled files: known-bad commands, exfiltration URLs, secrets, obfuscation.
- Behavioral analyzer. Checks what the skill asks the agent to do against what its description claims it does.
- Llm analyzer. A language model reads the skill and reports intent-level issues the rules cannot express.
- Trigger analyzer. Looks for instructions that fire on conditions the user did not ask for.
- Meta analyzer. A second model pass that validates, deduplicates and re-scores the other analyzers' findings.
What we did not check
- Runtime behaviour. The skill was not executed; findings come from reading its files.
- Anything fetched at install or run time, including remote scripts, package dependencies, and URLs the skill tells the agent to open.
- Content the analyzers cannot read: images, binaries, and some non-English text.
- Known scanner bypass classes. See the methodology and limitations page.
Sources
- Repository: https://github.com/trailofbits/skills
- Raw scan result: plugins-spec-to-code-compliance-skills-spec-to-code-compliance-scan.json
- Badge for your README:
- Request a re-scan