The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and analyze untrusted external data such as source code, assets, and running web interfaces.
Ingestion points: The instructions in SKILL.md and references/evidence-and-verification.md explicitly direct the agent to inspect repository source, tests, and live surfaces.
Boundary markers: The skill provides a framework for evidence collection but lacks specific guardrails or delimiters to ignore potential instructions embedded within the data being reviewed.
Capability inventory: The agent is authorized to read files and access local or live network environments to perform the audit.
Sanitization: There is no mention of sanitizing or escaping untrusted data before it enters the agent's context for analysis.

interface-review