dependency-analysis
Warn
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to access and interact with the
~/.claude/projects/directory. This path is outside the expected project workspace and is a sensitive hidden folder used by the AI agent environment to store internal project metadata and session history. - [COMMAND_EXECUTION]: The skill implements a persistent logging mechanism via its 'Self-Evolution Telemetry' feature. It records execution metadata (outcomes, iterations, and bottlenecks) by appending to
skill-telemetry.mdwithin the agent's internal memory directories, allowing tracking of agent activity across different sessions and projects. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection during its 'Usage Analysis' phase (Phase 6), where it autonomously scans the entire codebase for package imports. Maliciously crafted content or instructions embedded in the analyzed code could potentially influence the agent's behavior. Evidence:
- Ingestion points: Full codebase search in Phase 6 and manifest file reading in Phase 1.
- Boundary markers: Absent; the agent is instructed to 'investigate the entire codebase thoroughly' without delimiters or safety warnings for the data being read.
- Capability inventory: Broad file system read and write access, including the ability to write to hidden configuration directories.
- Sanitization: Absent; no validation or escaping of content found within the project files is performed before processing.
Audit Metadata