damage-prediction
Pass
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: Indirect prompt injection risk. The skill is designed to ingest and analyze untrusted data from the 'actual codebase' and 'damage tracking data models' as specified in Phase 1. It lacks boundary markers or explicit instructions to ignore embedded prompts within these external sources. The skill has the capability to write files to the 'docs/' directory, and no sanitization or escaping of external content is defined.
- [DATA_EXFILTRATION]: Local sensitive path access. The 'Self-evolution telemetry' section directs the agent to search for and write to '~/.claude/projects/', which is a hidden directory in the user's home folder used for agent state and project history. Accessing hidden platform configuration directories can expose metadata about the user's environment.
- [NO_CODE]: The skill consists solely of instructions within a markdown file and does not include any accompanying scripts or executable binaries.
Audit Metadata