The Agent Skills Directory

[Indirect Prompt Injection]: The skill analyzes untrusted document sets provided by the user, which could potentially contain hidden instructions designed to manipulate the agent's readiness score or findings. This is a common consideration for skills that process external content, and the risk is managed here by using objective metrics from scripts and a predefined rubric.
Ingestion points: Data enters the environment when scripts/analyze-exported-docs.py and scripts/detect-conflicts.py read files from the directory provided by the user.
Boundary markers: There are no specific delimiters used when the agent reviews the corpus, though the process is guided by structured templates like assets/readiness-report-template.md.
Capability inventory: The skill possesses the capability to execute local scripts and read files, but it does not have authorized tools for network access or system modification.
Sanitization: The input text is read as-is without specific filtering for prompt injection patterns, relying on the maker's oversight and the structured scoring rubric to maintain assessment integrity.- [Command Execution]: The skill instructions guide the agent to run local Python scripts to extract metadata and detect potential content conflicts. These scripts facilitate automated assessment and are a standard feature of the skill's design to provide objective data for the readiness report.

copilot-studio-knowledge-readiness