context-management
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to use standard shell commands such as git log, git diff, git stash, and cat to reconstruct session state and recover from context compaction events as documented in SKILL.md.
- [PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection by reading data from potentially untrusted files and git history.
- Ingestion points: Reads from .clarc/context.md, .claude/checkpoints.log, and git command outputs.
- Boundary markers: Instructions do not specify delimiters or markers to ignore embedded instructions in the ingested data.
- Capability inventory: Use of shell commands (git, cat) and session management commands (/checkpoint, /compact).
- Sanitization: No validation or sanitization routines are defined for the ingested context data.
Audit Metadata