Skip to content

Week 4 reading path

Read the notes and run the studio before the optional sources. The course examples are sufficient; no account, installation or external service is required.

  • Source 1 — instruction files: read the Custom instructions with AGENTS.md section of the Codex CLI documentation recorded in (tools/codex-cli; as_of: 2026-10-09; Source 1; Source 2; Source 3; Source 4; Source 5). Scope your reading to how instructions are discovered and combined. Compare that context mechanism with the studio's rule table; do not infer that instructions enforce filesystem permissions.
  • Source 2 — hooks: inspect the copied pre_commit.py and the Git event and exit-status contract described in the studio. This is a course-written synthetic example; record its input, blocked output and passing output rather than treating it as a vendor claim.
  • Source 3 — role context: read the subagents claim and its cited locator in (studies/humanlayer-context-engineering; as_of: 2026-10-09; Source 1). Scope the claim to limiting context and returning a compact hand-back. Treat the article's workflow as a proposal to test, not as evidence that a role is safe or effective.
  • Source 4 — skills as instructions: read the scope and progressive-disclosure claim in (tools/agent-skills; as_of: 2026-10-07; Source 1). Scope your comparison to a description, instructions and on-demand resources. A skill description does not grant a tool permission.

For each source, write one sentence naming the source section, the claim you used and the check in your studio that could support or challenge it. Factsheet cards are citations for this page; they are not published reading material.