The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones the Anti-Autoresearch repository from the author's GitHub account (https://github.com/wanshuiyin/Anti-Autoresearch.git) to the local directory ~/.claude/anti-autoresearch.
[REMOTE_CODE_EXECUTION]: The skill executes Python scripts (eval/run_eval.py and a dynamically resolved $GATE_HELPER script) from the cloned repository and local project directories. Execution is secured using a hardcoded SHA-1 commit hash (d8f510c49c29ccb5f98ecb1f8e397a7a27eb97c4) to ensure code integrity and prevent supply-chain attacks.
[COMMAND_EXECUTION]: The skill uses Bash to perform environment setup, repository management (git clone, checkout, reset, clean), and execution of audit tools. It requires broad tool permissions (Bash(*)) to manage the pinned repository state.
[PROMPT_INJECTION]: The skill processes untrusted research papers and PDFs, creating an attack surface for indirect prompt injection where adversarial content could influence the audit results.
Ingestion points: The skill accepts paper directories, PDF files, or ArXiv IDs as target arguments for analysis.
Boundary markers: The launcher delegates boundary management and prompt construction to the upstream audit tools within the cloned repository.
Capability inventory: The skill possesses extensive capabilities including arbitrary shell execution, file system modification, and script execution.
Sanitization: The process relies on a deterministic adjudication logic and hardcoded validation steps (eval gate) to ensure the integrity of the audit findings.

integrity-forensics