The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script codestable/tools/search-yaml.py using shell commands. The search pattern python codestable/tools/search-yaml.py --query "{关键词}" involves direct interpolation of user-supplied text into a shell command line. If the agent or the execution environment fails to properly escape shell metacharacters (e.g., ;, &, |, `) in the {关键词} placeholder, an attacker could execute arbitrary commands on the host system.
[INDIRECT_PROMPT_INJECTION]: Phase 1.5 of the workflow involves searching and reading existing decision documents from the codestable/compound/ directory. This creates a vulnerability surface where malicious instructions embedded in previously archived files could influence the agent's behavior during the archiving or review process.
Ingestion points: Files located in codestable/compound/ are read via the search-yaml.py tool.
Boundary markers: None identified. The skill does not specify using delimiters or instructions to ignore embedded commands when reading search results.
Capability inventory: The skill has the capability to write files to the project directory and execute local shell scripts.
Sanitization: No explicit sanitization or validation of the content of existing decision files is mentioned before processing.

cs-decide