The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python script (scripts/rdhs_query.py) to execute a local CLI tool named meegle. These executions are used to fetch real-time metrics and metadata from the Feishu Project space. The script uses subprocess.run with argument lists (avoiding shell=True) to minimize injection risks. All inputs to the CLI are either hardcoded identifiers or data retrieved from previous valid API calls.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from the Feishu Project platform, including product names, business lines, and role owner names. This data is interpolated into the agent's context for reporting. While this provides a surface for indirect prompt injection if the source data were maliciously crafted, the script implements strict schema validation and numeric checks on the retrieved data.
Ingestion points: Data is ingested via the meegle CLI output in JSON format in scripts/rdhs_query.py.
Boundary markers: The skill does not explicitly use delimiters or ignore-instructions for the external data, but it presents the data in a highly structured report format.
Capability inventory: The skill is limited to read-only CLI commands and does not have the capability to write to the system or execute arbitrary shell commands.
Sanitization: The script performs numeric validation (NaN check, range check) and JSON parsing, ensuring that only expected data types are processed.
[SAFE]: The skill emphasizes a "Read-Only" posture, explicitly forbidding automatic governance actions and avoiding the use of local disk caches to ensure data integrity and security regarding session credentials.

meegle-rdhs-diagnosis