The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive command execution, including running CLI tools, starting background servers, and executing git commands as part of the testing lifecycle.
[REMOTE_CODE_EXECUTION]: The skill generates executable shell scripts at runtime for 'PATH-based mocking' and executes them. This dynamic code generation and execution is a core feature used to simulate external dependencies.
[PROMPT_INJECTION]: The skill's 'spec-driven' architecture creates an attack surface where instructions embedded in untrusted E2E_TESTS.md files or project source code could influence the generated test code.
Ingestion points: E2E_TESTS.md files and project source code (analyzed via fp-init and fp-update-spec).
Boundary markers: Uses markdown headings and HTML comments for structure, but lacks explicit security boundaries to prevent instruction leakage from specifications to the code generator.
Capability inventory: Includes file system writes, process execution (runCommand), permission changes (makeExecutable), and network access via curl for verification.
Sanitization: No mention of sanitization or validation of specification content prior to code generation.

flightplanner