The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute a local script for validation during its operation cycle.
Evidence: Instructions in genes/gene_evolution.md and genes/GENES.md specify executing scripts/compile-validate.js during the 'Validate' phase of the gene evolution process.
[COMMAND_EXECUTION]: The instructions implement a self-modification behavior where the agent is directed to overwrite its own instructional files.
Evidence: The 'Solidify' step in genes/gene_evolution.md instructs the agent to 'write back Gene file' to persist updates derived from its execution history.
[PROMPT_INJECTION]: The skill features an 'evolution' process that utilizes untrusted external data to modify its own instructions, creating a surface for indirect prompt injection.
Ingestion points: The agent is instructed to 'Scan' compiler error logs and user feedback in genes/gene_evolution.md to identify failures and successes.
Boundary markers: While the skill uses a structured 'GEP' workflow, there are no instructions to sanitize ingested logs or to ignore embedded commands that might be present in compiler output or user feedback.
Capability inventory: The agent has the capability to write to local markdown files and execute scripts within the skill directory.
Sanitization: No explicit technical sanitization or validation of the input data is mentioned before it influences the 'Mutate' phase of the evolution cycle.

mpx2rn-gene