pruning-view
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local bash scripts to perform its primary function. Specifically, it runs
~/.claude/evolution/bin/evolve-prune.shand~/.claude/evolution/bin/evolve-regression.shto generate maintenance reports. - [INDIRECT_PROMPT_INJECTION]: The skill reads and summarizes generated markdown files (
reports/YYYY-MM-DD-prune.mdandreports/YYYY-MM-DD-regression.md). These reports are derived from candidate records and promoted learnings, which are untrusted data sources that could contain embedded instructions. - Ingestion points: Markdown report files located in the
reports/directory. - Boundary markers: No specific delimiters or warnings to ignore embedded instructions are specified when reading the report content.
- Capability inventory: The skill has the capability to execute shell scripts and propose system-level actions (like demoting learnings).
- Sanitization: There is no evidence of sanitization, escaping, or validation of the report content before it is processed and summarized for the user.
Audit Metadata