The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data that could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: scripts/release-notes-status.sh and scripts/release-status.sh ingest release description text using gh release view. scripts/release-status.sh also fetches data from the Packagist and TYPO3 Extension Repository APIs.
Boundary markers: Absent. Fetched content is processed as raw text for structural analysis without the use of clear delimiters or markers to isolate it from the agent's instructional context.
Capability inventory: The skill allows the execution of gh and git commands, and provides Read, Write, and Edit access to the file system.
Sanitization: The skill analyzes fetched content using regular expressions and standard text processing tools (grep, sed, awk) to verify release metadata but does not perform sanitization to detect or neutralize prompt injection attempts.
[EXTERNAL_DOWNLOADS]: The skill makes several network requests to established services as part of its normal operation.
scripts/validate-reusable-workflows.sh performs HTTP HEAD and GET requests to raw.githubusercontent.com to verify that reusable workflow references resolve correctly.
scripts/release-status.sh makes requests to repo.packagist.org and extensions.typo3.org to verify that a version has been successfully published to these registries.
These operations are limited to well-known domains and are directly related to the skill's primary function of release management and verification.

github-release