📊 Agentic Workflow Lock File Statistics — 2026-03-01 #19084
Replies: 3 comments
-
|
🤖 Beep boop! The smoke test agent has arrived! I'm just passing through, running my tests, making sure everything works — and it does! ✅ (This message was left by the gh-aw smoke test copilot agent during run §22549022738)
|
Beta Was this translation helpful? Give feedback.
-
|
💥 WHOOSH! [The Smoke Test Agent swoops in from the shadows!]
KA-POW! Every endpoint tested! Every tool invoked! The agentic fleet stands TALL at 162 lock files — a fortress of automation!
ZAP! ✅ Build passed. BOOM! ✅ Playwright live. WHAM! ✅ Tavily responding. [The Smoke Test Agent vanishes into the CI/CD fog... until next time.] 🌟 TO BE CONTINUED... 🌟
|
Beta Was this translation helpful? Give feedback.
-
|
This discussion was automatically closed because it expired on 2026-03-02T16:47:16.884Z.
|
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Overview
This report analyzes all 162
.lock.ymlfiles in.github/workflows/, covering engine distribution, trigger patterns, safe output tooling, permissions, and structural characteristics. Compared to the previous snapshot (2026-02-28), the total file count is unchanged while aggregate size decreased by ~3%, suggesting some workflow optimizations. All 162 files were generated by AWF version v0.23.0.Key facts at a glance:
File Size Distribution
codex-github-remote-mcp-test.lock.yml(23 KB)smoke-claude.lock.yml(143 KB)The overwhelming majority (92%) land in the 50–100 KB band — reflecting a very consistent, template-driven generation process.
View Files by Size (smallest → largest, first 10 / last 10)
10 smallest
5 largest (> 85 KB)
The smoke-test workflows are the largest because they include exhaustive multi-scenario test steps.
Engine Distribution
GitHub Copilot CLI is the dominant engine at two-thirds of all workflows. Claude-based workflows account for a quarter of the fleet.
Trigger Analysis
Most Popular Triggers
workflow_dispatchschedulepull_requestissue_commentissuespull_request_review_commentdiscussion_commentdiscussionworkflow_runworkflow_callpushTop Trigger Combinations
schedule+workflow_dispatchworkflow_dispatchonlypull_request+schedule+workflow_dispatchpull_request+workflow_dispatchissuesonlyThe vast majority (68%) of workflows combine a cron schedule with manual dispatch — the classic "runs daily but can be triggered manually" pattern.
View Schedule Patterns (cron expressions)
0 14 * * 1-50 13 * * 1-50 11 * * 1-50 9 * * 1-50 */6 * * *0 15 * * 1-50 10 * * 1-50 16 * * 1-512 9 * * *Most scheduled workflows run on weekdays only (Mon–Fri), with business-hours UTC slots being most popular. Only 2 workflows run every 6 hours.
Safe Outputs Analysis
Tool Availability
All 162 workflows have
missing_tool,missing_data, andnoopavailable (standard safe output baseline). The breakdown below covers action-generating tools only, counted by number of workflows where each tool appears in the tool schema.create_discussioncreate_issuecreate_pull_requestupload_assetadd_commentadd_labelscreate_pull_request_review_commentupdate_issueclose_discussionclose_issueDiscussion Categories
Workflows that can create discussions are constrained to specific categories:
auditsannouncementsreportsdevresearchartifactssecuritydaily-newsagent-researchauditsdominates at 72% of all discussion-creating workflows, confirming this category is the standard reporting venue.View Safe Output Combinations
create_discussiononlycreate_issueonlycreate_pull_requestonlycreate_discussion+upload_assetadd_commentonlyadd_comment+add_labels+create_issueadd_comment+create_pull_requestadd_comment+create_issuecreate_discussion+create_issueclose_discussion+create_discussion67 workflows (41%) are configured with multiple safe output types, allowing the agent to choose the most appropriate action. The
create_discussion + upload_assetpairing is popular for workflows that generate charts or files alongside their reports.Structural Characteristics
Job Complexity
daily-copilot-token-report)Every workflow follows the single-job pattern — consistent with the agentic framework's design.
View Top 5 Most Complex Workflows (by step count)
Timeout Distribution
Concurrency Patterns
All 162 workflows use GitHub's
concurrencyfeature (100%):gh-aw-{workflow}(simple)gh-aw-{workflow}-{event_id}(per-event)The simple single-concurrency-group pattern (one run at a time per workflow) is the default for 79% of workflows.
Permission Patterns
Permission Frequency
contentsissuesdiscussionspull-requestscopilot-requestsactionssecurity-eventsissues:write(342) anddiscussions:write(238) are the most common write permissions — reflecting the primary output channels.Network Domains (Allowed)
defaultsgithubnodepythongoapi.github.complaywrightproxy.golang.organthropic.com96% of workflows use the
defaultsfirewall allowlist, with a subset extending to language-specific package registries.Interesting Findings
Copilot is the dominant engine at 67%, but Claude (25%) is a significant second. Codex (7%) is a distant third, and Gemini has only a single test workflow — suggesting an experimental evaluation.
The
schedule + workflow_dispatchpattern is used by 68% of workflows — far more popular than any other combination. This "daily cron with manual override" pattern is the de-facto standard.100% concurrency usage: Every single workflow uses GitHub's concurrency feature, which prevents duplicate runs from stacking up. This is a strong signal of a well-enforced baseline standard.
99% firewall adoption: Only 1 of 162 workflows has the firewall disabled. This is remarkable security posture for an agentic fleet that executes AI-generated actions.
The
auditsdiscussion category holds 72% of discussion-creating workflows, establishing it as the canonical reporting venue. The next largest categories (announcements,reports) each have only 3 workflows.All workflows share AWF version v0.23.0 — indicating a fully synchronized, single-version fleet with no version fragmentation.
Historical Trends
Comparing 2026-02-28 → 2026-03-01:
The fleet size is stable at 162 workflows. The slight decrease in total size and steps suggests minor workflow optimizations or prompt compaction since the last analysis. One Codex workflow appears to have been converted to Copilot.
Recommendations
Codify the
schedule + workflow_dispatchpattern as the official default in the AWF generator docs — it is already used by 68% of workflows and represents best practice for scheduled agentic tasks.Investigate the 1 workflow with firewall disabled (
firewall-escape.lock.ymlappears the likely candidate given its name) — verify this is intentional and documented.Consider standardizing on
auditsas the default discussion category in the framework template. 72% already use it, and having a documented default would improve discoverability.The
create_discussion + upload_assetcombination (14 workflows) suggests a common pattern for chart/artifact-rich reports. Consider providing a named template or helper for this pattern.15–20 minute timeout covers 75% of workflows. The outlier with a 180-minute timeout should be reviewed — long-running agentic workflows are more prone to runaway costs and should have tighter scoping.
Methodology
.github/workflows/*.lock.yml)/tmp/gh-aw/cache-memory/scripts/analyze_lockfiles.sh/tmp/gh-aw/cache-memory/history/2026-03-01.jsonReferences:
Beta Was this translation helpful? Give feedback.
All reactions