The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest and analyze untrusted data from a local codebase, which could contain malicious instructions.
Ingestion points: Codebase files are scanned using LS, Grep, Glob, and Read in Step 4 (SKILL.md).
Boundary markers: The skill includes a dedicated "Security" section with a CRITICAL instruction forbidding the agent from reading or referencing .env files, API keys, tokens, or private keys.
Capability inventory: The skill utilizes Read, Grep, Glob, LS, Execute, and AskUser (SKILL.md).
Sanitization: There are no explicit instructions to sanitize or escape file content before it is interpolated into the final assessment report or the Twitter share link.
[EXTERNAL_DOWNLOADS]: External References to Trusted/Well-Known Sources. The skill references external URLs for legitimate functionality.
Twitter (X): Generates a sharing link using https://x.com/intent/tweet, a well-known service for social media integration.
Factory AI: Provides a link to https://docs.factory.ai for documentation, which is a vendor-controlled domain associated with the skill's author.

agent-repellent