gitingest
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary repository content into a text digest. The instructions explicitly warn the agent to treat this content as untrusted data, as it may contain instructions intended to influence the model's behavior (prompt injection).
- Evidence: SKILL.md advises treating digest contents as untrusted data; evals/evals.json includes a test case for refusing embedded instructions.
- [COMMAND_EXECUTION]: The provided
scripts/preflight.pyutility usessubprocess.runto executegitcommands (specificallygit rev-parse). This is used for verifying the integrity and identity of the local repository checkout against a pinned commit. The script uses a restricted environment and does not execute thegitingesttool itself. - Evidence: scripts/preflight.py lines 62-73 (git_output function).
- [CREDENTIALS_UNSAFE]: The skill identifies and warns against potential credential leakage. It notes that the underlying tool embeds tokens in the clone URL during operations and advises using fine-grained, read-only tokens in the environment instead of CLI arguments or hosted forms.
- Evidence: references/upstream.md (Safety findings) and SKILL.md (Gate every side effect).
- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the
gitingestpackage from PyPI and the cloning of repositories from GitHub. These actions are core to the skill's purpose and are gated by instructions requiring explicit user approval before execution. - Evidence: SKILL.md instructions for pip installation and remote ingestion.
Audit Metadata