The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes functionality to process untrusted external data through the ruview_claim_check {text} command, which evaluates reports, README sections, PR bodies, and model cards. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the processed data to manipulate the agent.
Ingestion points: External text processed via {text} parameter in SKILL.md.
Boundary markers: None identified to isolate untrusted data from the agent's instructions.
Capability inventory: Ability to execute shell scripts and Python scripts as defined in the verification workflows in SKILL.md.
Sanitization: No sanitization or escaping of external content is specified.
[COMMAND_EXECUTION]: The skill executes several local scripts to perform deterministic proofs and bundle verification.
Runs archive/v1/data/proof/verify.py via the ruview_verify command.
Runs scripts/generate-witness-bundle.sh to produce distribution artifacts.
Executes VERIFY.sh within a dynamically generated directory path (dist/witness-bundle-ADR028-*/) after the bundle is created.

verify