The Agent Skills Directory

[COMMAND_EXECUTION]: Executes local shell scripts provided by the target project as part of its verification process.
Evidence: Phase 2 instructions run bash hooks/session-start and other scripts referenced in hooks.json under PreToolUse or PostToolUse keys.
[COMMAND_EXECUTION]: Uses the Node.js runtime to dynamically load and inspect local JavaScript files from the project.
Evidence: The OpenCode test guide in references/platform-test-guides.md uses node -e to import plugin JS files directly.
[COMMAND_EXECUTION]: Performs file system modifications, including writing files outside the project root.
Evidence: Phase 1 instructions create a ../dev-marketplace/ directory and use shell redirection (cat > ...) to generate configuration files.
[PROMPT_INJECTION]: Vulnerable to indirect prompt injection via the project files it processes.
Ingestion points: Reads metadata and body content from SKILL.md and .md files in the skills/ and agents/ directories (SKILL.md, Phase 3).
Boundary markers: None identified to isolate project content from agent instructions.
Capability inventory: Can execute local scripts via bash and node, and write files to the local system.
Sanitization: No sanitization or validation of the extracted project metadata before it is used to generate the final test report or used in internal logic.

testing