tool-call-extract
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute Python scripts to search and extract data from local log files.
- Evidence:
SKILL.mdcontains instructions to runpython3 scripts/find-tool-calls.pyand references supplementary scripts in~/.pi/agent/skills/skill-call-extract/scripts/. - [INDIRECT_PROMPT_INJECTION]: The skill processes historical agent session logs, which can serve as an ingestion point for adversarial content encountered in past interactions.
- Ingestion points: The script
scripts/find-tool-calls.pyreads JSONL session logs from the~/.pi/agent/sessions/directory. - Boundary markers: While the data is extracted into a structured JSON format, the skill does not implement specific delimiters or instructions to ensure the agent disregards any commands found within the log content.
- Capability inventory: The skill has permission to read sensitive interaction logs and generate diagnostic artifacts on the local filesystem.
- Sanitization: No sanitization is performed on the raw tool arguments or result text extracted from the logs before they are presented to the agent for analysis.
Audit Metadata