The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading from and writing to external documentation files, creating a potential ingestion surface for indirect prompt injection.
Ingestion points: The skill ingests data by reading existing PLAN-*.md and ARCH-*.md files located within the docs/plans/** and docs/architecture/** directories (SKILL.md).
Boundary markers: No explicit boundary markers or instructions to disregard embedded content are specified in the skill's directives.
Capability inventory: The skill is restricted to Read and Write tool usage within specific documentation paths, with no network access or shell execution capabilities (SKILL.md).
Sanitization: The instructions do not mandate specific sanitization, filtering, or validation of the ingested file content before processing.

cost-model