The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents an indirect prompt injection surface because it ingests untrusted data that is used in shell command arguments.
Ingestion points: Target names (Step 3) and input PDB files (Troubleshooting Reference).
Boundary markers: No explicit delimiters or warnings to ignore embedded instructions are provided in the prompt templates.
Capability inventory: The skill has extensive capabilities including shell command execution (complexa design), file writes (write_manifest.py), and network operations (complexa download).
Sanitization: There is no evidence of sanitization or escaping for input strings before they are interpolated into CLI commands.
[COMMAND_EXECUTION]: The skill orchestrates several shell commands to run the four-stage design pipeline, including the complexa CLI tool and helper scripts like preflight.sh and write_manifest.py. These commands are essential for the primary scientific function of the skill.
[EXTERNAL_DOWNLOADS]: Fetches model weights, checkpoints, and community resources (AF2, RF3, Boltz2, ESMFold) from official vendor repositories and well-known services. These downloads are performed using the complexa download utility and are required for the pipeline's operation.

complexa-design