The Agent Skills Directory

[PROMPT_INJECTION]: Detection of injection patterns in templates/prd-example.md is a false positive; the phrase 'ignore previous instructions' is part of a functional requirement for an AI Support Agent described in a sample document, not an attempt to hijack the current agent's behavior.
[INDIRECT_PROMPT_INJECTION]: The skill can read existing files during its update/iteration phase, which provides a surface for indirect prompt injection if a user processes a malicious PRD file. However, this is inherent to the skill's legitimate purpose as a document editor. 1. Ingestion points: Reading previously saved or user-provided PRD files in SKILL.md (Step 6). 2. Boundary markers: None explicitly defined in the file reading logic. 3. Capability inventory: Uses Read, Write, Edit, and Bash tools to manage local files. 4. Sanitization: No specific sanitization of input file content is performed.
[COMMAND_EXECUTION]: The skill requests Bash access to perform file operations and saving as specified in the --out argument logic. This behavior is documented and aligns with the skill's stated purpose.

easy-prd