bindify
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local Python and Bash scripts (
blanche_gh_token.py,verify-worktree.sh,next_step.py,overview.py) to automate git workflows and reporting. These scripts usesubprocesscalls with argument lists (notshell=True), which prevents command injection vulnerabilities. The use ofopensslfor signing JWTs is handled correctly by passing paths as arguments. - [CREDENTIALS_UNSAFE]: The
blanche_gh_token.pyscript manages GitHub App private keys (PEM format) provided via environment variables (BLANCHE_GITHUB_APP_PRIVATE_KEYorBLANCHE_GITHUB_APP_PRIVATE_KEY_PATH). It processes these keys by writing them to a temporary file with restricted permissions (0600) and ensuring the file is deleted in afinallyblock after the operation is complete. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the repository (
wip-docs/folders and markdown artifacts) to coordinate updates and generate proposals. This constitutes an attack surface for indirect prompt injection. However, the skill provides extensive guardrails, such as mandatory summarization rules, a 'Clarify' gate that requires explicit user confirmation for intent and necessity, and instructions to stay strictly within the approved plan scope. - [EXTERNAL_DOWNLOADS]: The skill makes authenticated network requests to the official GitHub API (
api.github.com) to mint installation tokens for the Blanche bot and manage pull requests. These operations target a well-known service and are restricted to standard API functionality.
Audit Metadata