capture-knowledge

Warn

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill constructs and executes Python one-liners using the python3 -c flag by interpolating shell variables like $VAULT_PATH directly into the script string. This creates a risk of command injection where malicious input could execute arbitrary code.- [COMMAND_EXECUTION]: The skill executes local shell utilities such as grep, awk, and the obsidian CLI to perform vault discovery and file operations.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from the user's project files and conversation history.
  • Ingestion points: Reads files from the current repository and conversation context (SKILL.md).
  • Boundary markers: The instructions lack specific delimiters or instructions to ignore commands embedded in the processed text.
  • Capability inventory: The skill has filesystem write access and the ability to execute local shell and Python commands (SKILL.md).
  • Sanitization: No sanitization or validation of the distilled content is mentioned before it is written to the vault.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 10, 2026, 02:41 AM
Security Audit — agent-trust-hub — capture-knowledge