The Agent Skills Directory

[SAFE]: The skill implements a project management framework focused on documentation and process coordination. It restricts its operations to standard local file paths (e.g., docs/product/) and does not initiate network connections, access sensitive system credentials, or execute external scripts. All referenced actions are consistent with its stated purpose of feature launch management.- [PROMPT_INJECTION]: The skill handles data from external sources such as Product Requirement Documents (PRDs) and user feedback, which represents a surface for indirect prompt injection. However, the risk is negligible given the skill's restricted capabilities. 1. Ingestion points: The skill reads markdown files from docs/product/prd/ and processes data via the feedback-synthesis skill. 2. Boundary markers: The instructions do not define specific delimiters for separating untrusted data from agent instructions. 3. Capability inventory: Actions are limited to writing documentation and status files to the local file system and calling specific internal sub-skills. 4. Sanitization: No explicit content filtering or sanitization logic is implemented. The absence of high-risk capabilities like network exfiltration or shell access renders this surface safe.

feature-launch