The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes content from various external files (e.g., documents in company/ and .companify/) to perform its audit. These files act as untrusted inputs, creating a surface for indirect prompt injection where instructions hidden in the business artifacts could attempt to influence the agent's logic. Given the skill's restricted capabilities, the impact is minimal.
Ingestion points: The skill reads multiple artifacts from company/, .companify/assumptions.md, .companify/escolhas.md, .companify/metrics.md, and brand/strategy.md as documented in SKILL.md.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish between its own instructions and the content of the audited files.
Capability inventory: The skill is restricted to reading local files and writing a report to the .companify/reviews/ directory.
Sanitization: The skill does not perform sanitization or validation of the markdown content before processing.
[SAFE]: The skill follows best practices for its intended purpose, focusing on file-based auditing without requesting excessive permissions, network access, or dynamic code execution. All file paths are local and consistent with a project auditing workflow.

companify-audit