The Agent Skills Directory

[COMMAND_EXECUTION]: The validation logic in references/phase-assumptions.md instructs the agent to execute the shell command cat .planning/ROADMAP.md | grep -i "Phase ${PHASE}". The ${PHASE} variable is populated directly from the user-supplied $ARGUMENTS. If the agent executes this in a shell environment without sanitization, a malicious user could perform command injection by providing an argument such as 1; [malicious_command].
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from project files to generate its analysis.
Ingestion points: Content is read from .planning/STATE.md and .planning/ROADMAP.md as specified in SKILL.md.
Boundary markers: None are present to distinguish file content from instructions.
Capability inventory: The skill has access to file reading, shell command execution via the grep instruction, and file writing capabilities via the scripts/kata-lib.cjs utility (specifically the set-config command).
Sanitization: No sanitization or validation is performed on the phase number argument before it is interpolated into shell commands.

kata-list-phase-assumptions