The Agent Skills Directory

[PROMPT_INJECTION]: The skill architecture presents a vulnerability surface for indirect prompt injection due to its core function of generating platform logic from user-provided requirements.
Ingestion points: Untrusted data enters the agent context through the capability_gap_summary, target_use_cases, and expected_contract inputs defined in SKILL.md and contracts/skill-contract.yaml.
Boundary markers: Absent. The instructions do not define any delimiters or specific warnings for the agent to ignore instructions embedded within the user-provided integration requirements.
Capability inventory: Across SKILL.md, agents/openai.yaml, and contracts/skill-contract.yaml, the skill is granted capabilities to create, modify, and activate workflows and plugins via tools such as control.workflows.create, control.workflows.patch, control.workflows.activate, and control.plugins.create_plugin.
Sanitization: Absent. There is no evidence of sanitization, escaping, or strict schema validation performed on the natural language descriptions provided by the user before they are used to generate integration logic.

ascn-integrations