The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a mechanism to read and apply overrides from external files within the project directory.
Ingestion points: The skill reads .ai-factory/skill-context/aif-docs/SKILL.md (treated as a project-level override), .ai-factory/config.yaml, .ai-factory/DESCRIPTION.md, .ai-factory/ARCHITECTURE.md, and identifies other markdown files in the project root via Glob.
Boundary markers: There are no explicit boundary markers or instructions to disregard malicious prompts within the ingested data. The instructions explicitly state that 'the skill-context rule wins' when conflicts occur, which increases the risk of instruction hijacking if the files are malicious.
Capability inventory: The agent has access to Write, Edit, Bash(mkdir, npx, python), WebFetch, and WebSearch tools.
Sanitization: No sanitization or validation of the content within the skill-context or project files is described before it is used to influence the agent's behavior.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute mkdir -p docs-html during the HTML generation phase. This is an expected and restricted use of shell commands for creating necessary project directories.

aif-docs