The Agent Skills Directory

[COMMAND_EXECUTION]: The preamble section executes a local shell script 'check-update.sh' using a path relative to the skill directory ('../../check-update.sh'). This is a typical maintenance pattern for skills belonging to a larger repository or suite.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of reading and processing user-provided documentation. Ingestion points: Reads documentation from 'docs/02-方案设计/' and 'docs/01-需求调研/'. Boundary markers: Absent; there are no instructions to the agent to treat the input data as potentially containing instructions to be ignored. Capability inventory: The skill utilizes 'Write' to create output files, 'Bash' for local maintenance tasks, and 'Agent' for delegating sub-tasks. Sanitization: Absent; the skill extracts functional requirements directly from the source markdown files without sanitization or validation logic.

pm-proto