dev-onboarding

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted external data, creating a surface for indirect prompt injection.
  • Ingestion points: Reads meeting transcripts, proposals, MSAs, and POC outputs from a user-provided deal folder (SKILL.md Phase 2).
  • Boundary markers: The instructions for subagents lack explicit delimiters or 'ignore' warnings for potentially malicious content embedded in transcripts (SKILL.md Phase 2).
  • Capability inventory: The skill can execute shell commands via Bash and write to the local file system (SKILL.md Phase 5).
  • Sanitization: There is no evidence of sanitization or validation of ingested artifact content before it is used to generate documents.
  • [COMMAND_EXECUTION]: The skill executes a local shell script for document conversion.
  • Evidence: Runs bash ~/.claude/skills/dev-onboarding/scripts/build_outputs.sh to produce PDF and Word files (SKILL.md Phase 5).
  • Analysis: The script utilizes standard utilities (pandoc, soffice) and implements defensive practices such as quoting variables and using shell safety flags (set -euo pipefail) (scripts/build_outputs.sh).
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 12:01 AM
Security Audit — agent-trust-hub — dev-onboarding