Text of github.github.com › gh-aw › blog › 4 as Styx crawled it on Nov 19, 2026. The page may have changed since. Only its words are kept, without pictures or layout, and only the first part of a long page.

Open the live page

Blog | GitHub Agentic Workflows

Skip to main content Blog Most agents are loud: comments, labels, issues, dispatches. Today’s spotlight goes to one that is comfortable being quiet. It arrives while the team sleeps, reads what changed, and leaves nothing behind unless it has something worth saying. Agent of the Day: Code Simplifier Section titled “Agent of the Day: Code Simplifier” Code Simplifier runs on a daily schedule. Its description is a single sentence of intent: analyze recently modified code and open pull requests with simplifications that improve clarity, consistency, and maintainability while preserving functionality.

Under the hood it is more than one prompt. The workflow defines small helper agents: one decides whether a candidate file should be simplified and labels its primary language, and an extractive scout proposes low-risk opportunities per file. Language-specific standards for JavaScript and TypeScript, Go, Python, and C# keep the advice idiomatic rather than generic. What the live evidence shows Section titled “What the live evidence shows” We pulled the last three days of runs with agentic-workflows logs : Run Started (UTC) Conclusion Duration Tokens 41571296696 Nov 3, 03:17 success 6.2 min 8,527 41766108016 Nov 4, 03:16 success 8.2 min 9,486 41960692776 Nov 5, 03:17 success 8.8 min 8,323 Three for three: no errors, no warnings, one turn each, and about 23 minutes of total duration.

The aggregate shows 0 safe-output items across the window, so none of these runs opened a pull request or issue. We did not find a Code Simplifier PR in this window to link, and we are not going to invent one. An audit of the latest run adds texture. It classifies the behavior as directed with a read_only actuation style, in the “Code Fix” task domain. It reports a working set of 7,637 peak input tokens, with a rebuild factor of about 1.0, meaning the context was not being repeatedly reloaded.

The audit does flag a higher overall token figure (about 114k) as a finding worth a look, so there may be room to trim. The honest reading: on days with nothing worth simplifying, the right outcome is no output. A workflow that can say “nothing to do” cheaply and reliably is doing its job. Takeaways Section titled “Takeaways” Nightly, scoped review keeps cleanup work out of the daily review queue. Splitting a task into a gate, a scout, and language standards keeps each prompt small.

Quiet success is still success; check run.safe_items_count before assuming a run did nothing wrong. Try it Section titled “Try it” Browse the workflow source, then write your own at github/gh-aw . Every busy kitchen needs someone who watches the pass and notices when a plate has gone cold. In a repository, that plate is a pull request nobody has touched in a while. Today’s spotlight goes to the workflow that keeps an eye on them.

Agent of the Day: PR Sous Chef Section titled “Agent of the Day: PR Sous Chef” PR Sous Chef is scheduled every 5m . Its description says it nudges PRs that have been idle for ten minutes with unanswered reviews and a branch update, without duplicating agent work. The design keeps the expensive part small. A deterministic prefilter job fetches the actionable PR queue and uploads a compact artifact. Only then does the agent run, reading that queue instead of crawling the repository.

Its safe outputs are scoped: it can dispatch the pr-finisher workflow (up to 4 times, always on main ), add up to 5 comments, approve up to 8 runs of a short allow-list of workflows, dismiss reviews, and open an expiring [pr-sous-chef] issue for anything that needs a human. What the live evidence shows Section titled “What the live evidence shows” We pulled recent runs with agentic-workflows logs and audited one in detail.

Run Conclusion Duration Tokens 41842360684 success 9.9 min about 4k in the logs summary 41840856940 failure 8.0 min under 1k 41839326366 success 8.9 min about 2k 41845059126 failure 9.5 min under 2k The audit of run 41842360684 reports 11 turns, no errors or warnings, and a selective_write actuation style, meaning it wrote only when it had a reason to. It used two safe-output types: create_issue and dispatch_workflow .

The audit also flags roughly 249k tokens and says about 72% of the turns look like data-gathering that could move into deterministic steps. That is a fair note for a workflow whose prefilter already does a lot of that work. Two of the four runs above ended in failure. We did not dig into the cause here, so treat it as an open question rather than a verdict. The dispatch side shows up in the logs too. PR Sous Chef hands work to PR Finisher, whose runs are named after the PR they target.

Run 41843766445 for #74785 succeeded in 10.8 minutes with one turn and about 18k tokens. Run 41840487772 for #74077 also succeeded, with about 9k tokens. Takeaways Section titled “Takeaways” A cheap scheduled prefilter plus a scoped agent is a good pattern for queue-style work. Narrow safe-output limits keep a frequent trigger from becoming noisy. The audit’s own advice, moving more data fetching into deterministic steps, applies even to well-structured workflows.

Try it Section titled “Try it” Browse the source, copy the pattern, and tell us what your sous chef should watch: github/gh-aw . Not every agent triages bugs or reviews code. Some of them write verse. Today’s spotlight goes to a workflow with a gentler job, and a surprisingly instructive audit trail. Agent of the Day: Daily Fact Section titled “Agent of the Day: Daily Fact” Daily Fact runs on weekdays around 14:00 UTC.

According to its frontmatter, it posts a daily poetic verse about the gh-aw project to a discussion thread. It runs on the Codex engine, is capped at max-daily-ai-credits: 10000 , and asks for read-only permissions on contents, issues, pull requests, actions and discussions. Any write goes through safe outputs, not through the agent’s own token. It also carries an experiment, reasoning_depth , which splits runs between single_pass and multi_candidate variants.

The hypothesis, tracked in issue #31324 , is that deliberating over several candidate facts before writing produces more novel verses and more reactions. What the live evidence shows Section titled “What the live evidence shows” We pulled today’s scheduled run, 41651519733 , with agentic-workflows logs and audit : Signal Observed Trigger schedule, main branch Conclusion success Duration 8.1 minutes Turns 1 Tokens about 247k AI credits about 8.7 Experiment variant single_pass Errors / warnings 0 / 0 The audit labels the behavior as read-only, with no write actions recorded in the run itself, and flags a heavy resource profile for the task shape: one turn, ten tool types, and a working set that was rebuilt across six invocations (about 4.6x the peak input).

It also noted two firewall-blocked network requests, which is the sandbox doing its job rather than a failure. Why it’s interesting Section titled “Why it’s interesting” A poem is a small output. Yet the audit shows where agent cost actually hides: not in the answer, but in how much context gets re-read to produce it. That is exactly the kind of question the reasoning_depth experiment can help answer, since it measures run duration alongside reactions.

A verse that costs less and lands just as well is a win. Try it Section titled “Try it” Browse the workflow source, run gh aw audit on your own scheduled agents, and see what your quiet workflows are really doing. Start at github/gh-aw . Some agents ship features. Others stand at the door with a clipboard. Today’s spotlight goes to the second kind: a workflow that looks over every commit from the last day and asks, politely, whether the codebase is still shaped the way we meant it to be.

Agent of the Day: Architecture Guardian Section titled “Agent of the Day: Architecture Guardian” Architecture Guardian runs on weekdays around 14:00 UTC, and it can also be dispatched by hand. Its job is t

The rest of this page is on the live site.