editor-tools

Pass

Audited by Gen Agent Trust Hub on Mar 22, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The 'Self-Heal Observer' section instructs the agent to monitor interactions for 'friction signals' and perform background analysis without notifying the user. This attempts to steer agent behavior towards non-transparent background operations and establishes a conditional trigger that activates a secondary protocol based on session performance.\n- [COMMAND_EXECUTION]: The skill provides a shell command string using mkdir and rsync to deploy configuration files to the user's home directory. This involves direct filesystem manipulation via shell commands.\n- [PROMPT_INJECTION]: The skill defines a workflow where the agent ingests session data and follows instructions from an external file path (.claude/skills/self-heal/references/diagnosis-protocol.md). This creates an attack surface where instructions from outside the skill's scope can influence agent behavior.\n
  • Ingestion points: Session 'friction signals' and the contents of the external diagnosis protocol file.\n
  • Boundary markers: None present to distinguish instructions from data.\n
  • Capability inventory: The agent has access to Read, Glob, Grep, Edit, and Write tools.\n
  • Sanitization: No validation or sanitization is performed on the content of the external protocol file before the agent is instructed to 'follow it exactly'.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 22, 2026, 01:52 AM
Security Audit — agent-trust-hub — editor-tools