The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input (product ideas) during its interactive sessions to generate structured documentation such as the Opportunity Brief and Context Document. This data is intended for consumption by downstream agents, presenting a surface for potential indirect prompt injection.
Ingestion points: User descriptions of ideas and problems in Phase 0 and Phase 1 as described in SKILL.md.
Boundary markers: The skill uses a structured YAML schema (defined in templates/product-context-schema.yaml) to delimit different sections of the product definition.
Capability inventory: The skill has the ability to write to the local file system (product-context.yaml, product/index.yaml) and perform git commits.
Sanitization: The skill performs YAML syntax validation using js-yaml, but does not implement semantic filtering for instructions within the generated text.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to use npx to execute validation tools, which may result in downloading packages from the npm registry.
Evidence: Use of npx js-yaml and npx tsx in SKILL.md and references/yaml-guardrails.md.
Context: These tools are standard for development environments and target the well-known npm registry.
[COMMAND_EXECUTION]: The skill utilizes local shell commands for environmental checks, file validation, and project management.
Evidence: Execution of ls to detect project modes, git add/commit for version control, and npx for tool execution as specified in SKILL.md.

aep-envision