atlas-resume
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes
python3andgitcommands to manage task resumption and cleanup operations. - [DYNAMIC_EXECUTION]: It utilizes
python3 -cto execute Python code that imports internal modules from the skill's directory structure. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the user's current working directory to restore agent state.
- Ingestion points: Reads state and plan data from
.atlas/state.jsonand.atlas/plan.dag.jsonin the current working directory. - Boundary markers: Absent. The skill instructions do not define delimiters or validation for data recovered from disk.
- Capability inventory: Includes file system writes, file deletions through
git, and execution of local Python scripts. - Sanitization: Absent. The instructions assume the integrity of the ledger files found in the user's project.
Audit Metadata