archfit-eval
Pass
Audited by Gen Agent Trust Hub on Aug 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands such as
make build,make test, andpython3 scripts/eval/corpus_sweep.py. These commands are used to prepare the environment and run automated evaluations. This grants the agent shell capability, which is standard for a development-oriented evaluation skill.- [PROMPT_INJECTION]: The skill processes untrusted data from external repositories in the test corpus, which presents a surface for indirect prompt injection attacks. - Ingestion points: Files within target repositories being evaluated (e.g., via the
--rootflag incorpus_sweep.py). - Boundary markers: Absent; the instructions do not specify the use of delimiters or 'ignore' instructions for the data being processed.
- Capability inventory: Shell access, local binary execution (
.bin/archfit), build tool execution (make), and script execution (python3). - Sanitization: No sanitization or filtering logic is defined in the skill instructions for handling content from the target repositories.
Audit Metadata