The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script python .codestable/tools/search-yaml.py to search through existing documentation. It uses placeholders like {组件名} and {主要 tag} to construct command arguments, which involves interpolating user-provided or context-derived strings into a shell command.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks (Category 8) because it processes untrusted data and reads it back into the agent context.
Ingestion points: Summaries are generated from conversation context and user input during Phase 1 and Phase 2. Existing documents are read back during Phase 1.5.
Boundary markers: The instructions do not define clear delimiters or include 'ignore embedded instructions' warnings when reading or writing documentation content.
Capability inventory: The skill has the capability to write files to the .codestable/compound/ directory and execute a local Python search tool.
Sanitization: There is no explicit evidence of sanitization or validation of the content gathered from the user before it is written to files or read back into the prompt.

cs-learn