The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its handling of untrusted user input.
Ingestion points: The problem statement provided via $ARGUMENTS in SKILL.md is ingested and used to drive the entire agent workflow.
Boundary markers: The instructions in rules/divergence.md interpolate the "selected framing" (derived from user input) into subagent prompts without explicit boundary delimiters or "ignore embedded instructions" warnings.
Capability inventory: The skill can spawn multiple subagents (generators, judges) and perform persistent memory writes through memory.write.
Sanitization: There is no evidence of sanitization or structural validation of the problem statement content before it is processed by subagents.
[COMMAND_EXECUTION]: The skill references local scripts that are executed during diagnostic and validation phases.
Evidence: rules/diagnostic-surface.md declares node scripts/eval/l1.mjs and bash scripts/sync-symlinks.sh as repository validators.
Context: While these are standard development tools, they represent shell and runtime execution within the local environment.

ideate