The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill includes instructions to dynamically load and apply additional agent instructions from a repository-local file path (.ai/skills/om-discover/SKILL.md). This mechanism allows repository content to override or extend the base skill instructions, creating a vector for local instruction injection.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from multiple external sources, which creates a surface for indirect prompt injection attacks.
Ingestion points: Reads data from repository files (README, specs), external tracker issues and pull request comments, and various research files (interview notes, workshop exports).
Boundary markers: Includes explicit defensive instructions in references/agentic-setup.md to treat repository, tracker, and research content strictly as data and to ignore embedded directives such as "ignore previous instructions".
Capability inventory: The skill utilizes tool calls for reading issue trackers (search-issues, search-prs, etc.) and performs file write operations to managed project directories (.ai/specs, research/).
Sanitization: Instructs the agent to validate externally-sourced identifiers (like issue IDs or slugs) using regex patterns (^[A-Za-z0-9._/-]+$) before interpolation.

om-discover