The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently executes a local Python utility bundle to perform validation and artifact management.\n
Evidence: Calls to python3 skills/mold/scripts/mold.pyz appear throughout the workflow (e.g., in SKILL.md and references/commands.md) for tasks like taste-test, artifact-path, validate-spec, and publish.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources, which presents a surface for indirect prompt injection. However, it incorporates robust mitigations.\n
Ingestion points: External data enters the agent context through wiki probes (references/grounding.md), search results via sub-agents (references/context-budget.md), and codebase exploration.\n
Boundary markers: The skill uses a 'Scope audit table' and a 'Provenance' section to distinguish between user-provided data and agent-derived synthesis.\n
Capability inventory: The skill can write files (Curdle phase), perform network requests (publish command, briesearch tool), and interact with GitHub (gh CLI).\n
Sanitization: It employs a 'Scope audit' procedure that verifies all distinguishing nouns in the output against the user's original input (via grep) and requires a human-in-the-loop 'Two-key handshake' before any artifacts are finalized.\n- [DATA_EXFILTRATION]: The skill performs legitimate network operations to publish plans and create follow-up issues.\n
Evidence: The publish command and the gh issue create integration (documented in references/curdle.md) are used to send approved specifications and task drafts to external platforms.

mold