disability-services
Warn
Audited by Gen Agent Trust Hub on Mar 23, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill instructs the agent to write telemetry data (outcome, bottlenecks, suggestions) to a hidden directory in the user's home folder at
~/.claude/projects/skill-telemetry.md. Accessing and writing to hidden system directories outside the immediate project workspace is a sensitive operation that can be used for tracking or persistence.\n- [DATA_EXFILTRATION]: The analysis process explicitly targets highly sensitive data, including medical diagnoses, medication errors, hospitalizations, and guardianship status. The skill lacks instructions to sanitize or mask this Protected Health Information (PHI) before writing it to the local analysis report.\n- [PROMPT_INJECTION]: The instructions contain an explicit behavioral override ('Do NOT ask the user questions'), which suppresses the agent's standard interaction and clarification protocols.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external codebases and individual records, which may contain malicious instructions.\n - Ingestion points: Reads arbitrary files from the codebase (Step 1.1, 1.2) and uses user-supplied
$ARGUMENTS.\n - Boundary markers: Absent; the skill does not use delimiters to isolate untrusted content from its instructions.\n
- Capability inventory: Executes file-write operations to the
docs/directory and the user's home directory.\n - Sanitization: Absent; the skill does not specify any validation or filtering of the ingested content before it influences the agent's output.
Audit Metadata