The Agent Skills Directory

[DATA_EXFILTRATION]: The instructions in references/reverse-engineer/analysis-engine.md (Section 7) direct the agent to parse .env files to identify external API services. Environment files are highly sensitive as they are the standard location for storing application secrets, database credentials, and API keys. Accessing these files poses a significant risk of exposing sensitive information to the model's context.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze arbitrary codebases (Stage 2 in SKILL.md). It does not implement boundary markers or instructions to treat the ingested code as untrusted data, which could allow malicious instructions embedded in the analyzed code to influence the agent's behavior.
Ingestion points: Full project codebase including route files, configuration files, and sensitive .env files as documented in analysis-engine.md.
Boundary markers: Absent; there are no instructions to isolate or ignore instructions embedded within the codebase being analyzed.
Capability inventory: File system read, glob, and write operations used to process code and generate documentation files.
Sanitization: The skill does not mention sanitizing or escaping data extracted from the codebase before it is used in documentation generation or subsequent reasoning steps.

vibe-coding-reverse-engineer