The Agent Skills Directory

[SAFE]: The skill content is purely instructional, establishing a 'debug' discipline (Reproduce, Isolate, Hypothesize, Fix, Verify). It does not include any executable scripts, shell commands, or network operations. The provided evaluation files (README.md and cases.yaml) are for testing the agent's behavior and do not introduce security risks.
[INDIRECT_PROMPT_INJECTION]: The skill defines an interaction surface by reading technical_level from 02-DOCS/wiki/harness/user-profile.md and model routing configuration from 02-DOCS/wiki/sdd/config.yaml.
Ingestion points: SKILL.md (lines 35, 45) instructs the agent to read local project files for user preferences and session configuration.
Boundary markers: None present.
Capability inventory: The skill coordinates other stack skills (FastAPI, Go, Next.js, etc.) which possess testing and debugging capabilities.
Sanitization: None present.
Assessment: The risk is negligible as the ingested data is used strictly for technical register adjustment (vocabulary analogies) and internal model tier routing, rather than being treated as executable instructions.

debug