[research] Overview of docs improver agents

[mnkiefer]

I want to analyze all agentic workflows in this repository that contribute to our docs improvements. Give me a table of their names and descriptions and a brief summary of their percentage contributions so far, what worked and what didn't.

[mnkiefer]

/scout history=full

[github-actions]

🔍 Scout Research Report

Triggered by @mnkiefer

Executive Summary

This repository has 9 agentic workflows that contribute to documentation improvements. Of those, 5 generate commits directly (56 total [docs]-prefixed commits found in history), while 4 primarily output to GitHub Discussions, Issues, or assets. Documentation Unbloat is the dominant contributor (~43% of all docs commits), followed by the Daily Documentation Updater (~20%) and Developer Documentation Consolidator (~14%).

---

Agent Overview

Workflow Name	File	Description	Output Type	Schedule
Daily Documentation Updater	daily-doc-updater	Automatically reviews and updates documentation for accuracy and completeness	PRs (auto-merge)	Daily
Daily Documentation Healer	daily-doc-healer	Self-healing companion to DDU — detects gaps missed by the updater and proposes corrections	PRs / Issues	Daily
Documentation Unbloat	unbloat-docs	Reduces verbosity and removes bloat while maintaining clarity	PRs (auto-merge, draft)	Daily + on PR comment
Developer Documentation Consolidator	developer-docs-consolidator	Consolidates and organizes developer docs from multiple sources into a unified knowledge base	PRs + Discussions	Daily
Glossary Maintainer	glossary-maintainer	Maintains and updates the documentation glossary based on codebase changes	PRs	Weekdays (10am UTC)
Claude Code User Docs Review	claude-code-user-docs-review	Reviews docs from the perspective of a Claude Code user (non-Copilot)	Discussions (audits)	Daily (8am UTC)
Multi-Device Docs Tester	daily-multi-device-docs-tester	Tests documentation site functionality and responsive design across mobile/tablet/desktop	Issues + assets	Daily
Documentation Noob Tester	docs-noob-tester	Tests docs as a new user would — finds confusing or broken getting-started steps	Discussions (audits) + assets	Daily
Technical Doc Writer	technical-doc-writer	Reviews and improves technical documentation for a given topic	PRs + comments + assets	Manual (workflow_dispatch)

---

Contribution Breakdown (Commit-Based)

Of 56 [docs]-prefixed commits in the full repository history:

Agent	Commits	% of [docs] commits	Notes
Documentation Unbloat	~24	~43%	Includes unbloat, remove bloat, reduce bloat, streamline, remove repetitive, redundant content
Daily Documentation Updater	11	~20%	Consistent [docs] Update documentation for features from YYYY-MM-DD pattern
Developer Docs Consolidator	~8	~14%	Versioned dev.md progression (v2.x → v3.7), tone fixes, code fence repairs
Glossary Maintainer	5	~9%	[docs] Update glossary — daily and weekly full scans
Daily Documentation Healer	4	~7%	[docs] Self-healing documentation fixes — runs but fewer accepted PRs
Claude Code User Docs Review	0	—	Outputs to Discussions only (no direct commits)
Multi-Device Docs Tester	~12 smoke test files	—	Creates issues + asset screenshots; smoke test files appear in branches
Documentation Noob Tester	0	—	Outputs to Discussions + uploads assets only
Technical Doc Writer	0 (in history)	—	Manual-only trigger; no scheduled commit history found

---

What Worked Well

• Documentation Unbloat is the clear leader in raw commit volume. Measurable percentage reductions are explicitly stated in commit messages (e.g., "24% reduction", "56% reduction", "72% reduction"), indicating the agent is quantifying its own impact. It handles both scheduled daily runs and on-demand PR comment triggers.

• Daily Documentation Updater shows consistent, predictable coverage — one PR per day tracking new feature documentation. The auto-merge configuration means accepted PRs land quickly.

• Developer Documentation Consolidator demonstrates a well-structured versioning approach (dev.md v2.0 → v3.7), maintaining a single canonical developer specification file. The combination of Discussions (for audits) + PRs (for actual changes) is a good pattern.

• Glossary Maintainer is steady and targeted — both a daily scan mode and a weekly full scan mode, with wiki memory enabled for context continuity across runs.

• Daily Documentation Healer serves as a safety net for the Updater. Even though only 4 commits are attributed to it, its role as a gap-detector is architecturally sound.

What Didn't Work / Gaps

• Daily Documentation Healer has very few merged contributions relative to its daily frequency. The gap between "runs daily" and "only 4 commits" suggests either low yield, high PR rejection rate, or that it surfaces issues as GitHub Issues rather than committed fixes.

• Technical Doc Writer has no commit history from scheduled runs — it's manual-only (workflow_dispatch). This limits its impact to one-off, human-initiated invocations. No evidence of regular use was found in the git log.

• Multi-Device Docs Tester creates smoke test files and GitHub Issues, but no documentation improvements — it is diagnostic, not corrective. Its impact on actual docs content depends on humans acting on the issues it creates.

• Documentation Noob Tester and Claude Code User Docs Review are similarly diagnostic: they surface problems in Discussions, but there's no evidence of automated follow-through pipelines that convert discussion findings into PRs. The loop may be manually closed.

• No agent currently tracks cross-agent coordination: Unbloat and Healer could theoretically conflict if operating on the same files simultaneously (both run daily). No deduplication or scheduling offset is visible in the workflow configs.

---

Suggested Next Steps

1. Activate Technical Doc Writer on a schedule (e.g., weekly) for recurring high-priority topics, rather than relying solely on manual triggers.
2. Close the loop on diagnostic agents: Create a downstream workflow that reads Discussions from Noob Tester and Claude Code Review, then opens tracked Issues or auto-assigns to the Healer.
3. Track Healer yield rate: Instrument the Healer to report how many PRs it creates vs. how many are merged — low yield may indicate prompt tuning is needed.
4. Add conflict prevention: Ensure Unbloat and Healer don't target the same files in the same window (e.g., stagger their cron times or add file-lock logic).

References:

• §22757597317

🔭 Intelligence gathered by Scout · ◷