self-healing-context

Pass

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted local data.
  • Ingestion points: Reads content from ground-truth.md, crm/companies.md, crm/opportunities.md, crm/outreach.md, and weekly logs in plans/.
  • Boundary markers: The instructions do not specify any delimiters or safety prompts to prevent instructions within these files from overriding the agent's logic.
  • Capability inventory: The agent has access to Write and Bash tools, which could be misused if malicious instructions are successfully injected via the data files.
  • Sanitization: There is no mention of sanitizing or validating the input data before it is parsed and used to generate the reconciliation report.
  • [DATA_EXFILTRATION]: The skill accesses sensitive business and career data from the local environment. However, there are no network operations or external communication patterns detected, meaning data remains local to the workspace.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 29, 2026, 05:16 AM
Security Audit — agent-trust-hub — self-healing-context