The Agent Skills Directory

[SAFE]: The skill provides structured domain knowledge for proposal writing and contains no active scripts, shell commands, or binary executables. The entire capability is instructional in nature.- [EXTERNAL_DOWNLOADS]: The documentation references external finance and accounting doctrine hosted on the author's official GitHub account (github.com/peterbamuhigire/chwezi-accounting-doctrine). These resources are informative and consistent with the author's stated identity.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external documents such as solicitations, Terms of Reference (ToR), and project briefs. While this creates a surface for potential data-driven instruction overrides, the skill includes robust safeguards to mitigate this risk. 1. Ingestion points: External buyer solicitations, ToRs, and profiles are read by the agent as specified in the 'Required Inputs' tables. 2. Boundary markers: The instructions include explicit stop conditions, conformance checks, and 'Anti-Pattern' warnings to ensure the agent remains compliant with the framework logic. 3. Capability inventory: The skill is limited to reading workspace files and preparing routing guidance; it lacks the ability to execute code or perform network operations. 4. Sanitization: The skill mandates evidence recording and source verification, requiring the agent to mark unverified data as 'not assessed' rather than accepting it as fact.

sectors