The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through its context-loading feature. This is a standard functional requirement for localized content generation.
Ingestion points: The skill reads creator-specific data from a local HOOKS.md file located within the project hierarchy via the scripts/load-context.mjs script.
Boundary markers: There are no explicit delimiters defined for the content of the HOOKS.md file, though the skill expects a specific markdown structure as defined in the init command.
Capability inventory: The skill is restricted to text generation, auditing, and local file discovery. It lacks network access, arbitrary code execution, and high-privilege system access.
Sanitization: The skill does not perform explicit sanitization of the input file content, relying on the agent's core safety protocols and the predefined instructional templates.
[COMMAND_EXECUTION]: The skill invokes a local Node.js script (scripts/load-context.mjs) to automate the discovery of user-provided context. The script utilizes standard Node.js file system and path modules for identifying file locations and does not perform network operations or execute external shell commands.

hooksmith