The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies an attack surface for indirect prompt injection as it ingests untrusted project source code and configuration files.
Ingestion points: Reads local files such as package.json, source files via globbing (app/**/route.ts, pages/**/*.{tsx,jsx}), and other project-specific manifests.
Boundary markers: The instructions do not define specific delimiters or guardrails to prevent the agent from obeying instructions embedded in the analyzed code.
Capability inventory: The skill is restricted to filesystem operations, specifically reading project files and writing documentation to a .context/ directory.
Sanitization: No explicit sanitization or filtering of processed source code is implemented before analysis.
[COMMAND_EXECUTION]: The skill uses git diff in its incremental update mode to determine which files have changed since the last discovery. This command is executed locally and is standard for development tools interacting with repositories.

workflow-discover