ar-resume
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a Python script (
setup_experiment.py) located within the skill's own directory and uses standard shell commands (git checkout,git log,cat) to list and retrieve local experiment state. These commands are executed to handle branch management and file reading within the project scope. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from local experiment files to summarize progress for the user, representing a potential but standard ingestion surface.
- Ingestion points: Reads content from
.autoresearch/{domain}/{name}/(config.cfg, program.md, results.tsv) and Git logs to generate an experiment summary. - Boundary markers: The skill does not explicitly define markers but uses the data to inform the user of status rather than executing it as primary instructions.
- Capability inventory: File reading, Git branch manipulation, and execution of internal setup scripts.
- Sanitization: The skill summarizes the data for human review before suggesting next steps (/ar:run or /ar:loop).
Audit Metadata