The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and synthesize data from external, potentially untrusted sources including issue trackers, pull request comments, and team chat messages.
Ingestion points: Data is collected via tools like git blame, git log --follow -p, and gh pr view, as well as category-mapped MCPs for chat, observability, and analytics as described in SKILL.md Step 2.
Boundary markers: The skill incorporates a detailed epistemic guide (references/epistemics.md) that mandates the use of confidence tiers and specific hedging language to separate raw evidence from agent inference, serving as a cognitive boundary.
Capability inventory: The skill utilizes read-only capabilities to query source control, documentation repositories, and communication platforms.
Sanitization: Instructions include a mandatory 'Calibration Check' to verify citations and ensure that the agent does not treat code shape as proof of intent before finalizing the response.
[PROMPT_INJECTION]: The SKILL.md and TESTS.md files contain instructions using 'OVERRIDE' terminology (e.g., <HARD-GATE> OVERRIDE SHIP, OVERRIDE: user required the skill) which direct the agent to disregard previous baseline judgments or automated checks regarding the skill's necessity.

why