References: Agents in the pipeline
Source material
Section titled “Source material”Every flag name and mode description in this lesson was verified against the public Claude Code documentation; short quotes are verbatim from the pages below. The worked example (Priya’s team, the booking app, the nightly generator, and the pre-merge reviewer) is Clawdemy’s own.
- Run Claude Code programmatically, Claude Code documentation. The lesson’s primary source: the print flag for non-interactive runs, piping through standard input and output, the output format options (“structured JSON with result, session ID, and metadata” for the JSON option), the JSON schema flag for fixed-shape output, pre-approving tools, permission modes for CI, continuing conversations across runs, and bare mode (“Bare mode is useful for CI and scripts where you need the same result on every machine”).
- CLI reference, Claude Code documentation. The canonical flag list behind this lesson: print, output and input formats, the turn limit (which exits with an error at the cap), the per-run dollar budget, permission modes, and the system prompt flags.
- Claude Code GitHub Actions, Claude Code documentation. Anthropic’s managed CI integration, and the source of “Claude respects your CLAUDE.md guidelines and existing code patterns.” Its best-practices section recommends putting review criteria and project rules in project memory, and its cost guidance (turn caps, timeouts, concurrency limits) matches this lesson’s budget discipline.
- Common workflows, Claude Code documentation. The pipe-Claude-into-scripts pattern and the options for running work on a schedule, from CI pipelines to managed routines.
Going deeper
Section titled “Going deeper”- Claude Code in Action, Anthropic Academy. Free, self-paced official course on integrating Claude Code into a development workflow, a natural hands-on companion to this lesson’s pipeline patterns.
- Building Effective AI Agents, Anthropic engineering blog. The simplicity principle this lesson leans on when it insists both pipeline jobs stay drawn routes rather than free-roaming agents.
On this site
Section titled “On this site”- CLAUDE.md at team scale (this track, lesson 2). The configuration layer the CI agent inherits: project memory, scoped rules files, and the audience question this lesson reuses.
- Anatomy of an AI Agent Team (Track 25). Independent review as an organizational principle, studied in a working multi-agent system; this lesson turns the same principle into pipeline mechanics.
Adjacent topics
Section titled “Adjacent topics”- The next lesson is the capstone: design, build, and defend a small agentic system of your own, with every trade-off from this track on the table.
- If the budget discipline interested you most, the GitHub Actions page’s cost section is the practical follow-up: turn caps, workflow timeouts, and concurrency controls in a real CI setting.