docs: lead with Auditable Workspaces category framing by Koukyosyumei · Pull Request #244 · h5i-dev/h5i

Koukyosyumei · 2026-06-30T12:48:19Z

No description provided.

Homepage + pitch now make 'Auditable Workspaces for AI Coding Agents' the H1/category, demote the multi-agent team to the killer demo, surface single-agent value first, and broaden the problem into a single-agent vs multi-agent split. Added a 5-card core-capabilities section, moved the multilingual proverbs out of the hero, and removed dash punctuation from all visible copy on both pages.

.hero-inner was a flex item with no explicit width, so it shrank to fit its widest child (the nowrap title). The embedded h5i serve dashboard, a block child, inherited that width — coupling the embed size to the headline length. Set width:100% and raise max-width to 980px so the embed renders at its own designed max-width regardless of the title.

Rework the core-capabilities cards so each one explains a real h5i command family: capture (provenance + token-lean output), recall (read the AI history), env (sandboxed worktrees), team (agent teams + neutral verifier), audit (deterministic risk assessment), and share (evidence travels with the branch). Verbs verified against the live CLI help.

Trim each 'What h5i is' card to one or two short sentences.

Move the verbs that live under a noun into that noun's section instead of the catch-all 'h5i objects' page: - h5i capture run + Structured output -> under h5i capture - h5i recall object / objects + h5i recall rm -> under h5i recall - h5i objects keeps only its store-maintenance verbs (gc/pin/fsck, push/pull, filters/trust, setup), with an intro pointing back to the capture/recall entry points Also move the h5i hook section up next to h5i capture (hooks drive the automatic prompt capture). Update the hand-written TOC and regenerate docs/manual/index.html. No new broken anchors (verified).

…h5i objects' Drop the parenthetical and update the slug to #h5i-objects across the TOC and every cross-reference; regenerate docs/manual/index.html.

…rentheticals 'h5i env (isolated agent sandboxes)' -> 'h5i env' and 'h5i team (auditable agent ensembles)' -> 'h5i team'; update slugs to #h5i-env / #h5i-team across the TOC and all cross-references (also repairs the appendix 'Sandbox / environments' self-anchor). Regenerate docs/manual/index.html.

Koukyosyumei · 2026-06-30T12:48:52Z

🪙 78% AI-authored · 4 files · prompt 48/100

7 AI commits

Important

🎯 Goal: Restructure docs site: make 'Auditable Workspaces for AI Coding Agents' the H1/category on homepage + pitch, demote multi-agent team to killer-demo, broaden problem section, single-agent value first

7 / 9	4	1 : 1.6	1 / 6
_{AI commits}	_{files touched}	_{READ : EDIT}	_{THINKs / ops}

💬 The ask

"Do you think we can further improve the speed without any security violation? Can you then review #239? Let's further restructure / improve the website in ./doc. "うん、“Auditable Workspace” 寄りにしたほうがいいと…"

📍 What shipped

✓ docs: map the 'What h5i is' cards to the six command families
✓ docs: tighten the six core-capability card descriptions
✓ docs(manual): regroup the CLI reference by command family
✓ docs(manual): rename the 'h5i objects (token reduction)' section to 'h5i objects'
✓ docs(manual): rename 'h5i env'/'h5i team' section headings to drop parentheticals

Note

🪴 Prompt maturity: 48/100 · developing · 7 prompts scored (100% of AI commits)
🔧 Recurring weak spots: no acceptance criteria, weak context, too short.
Heuristic signal of prompt craft — not a developer rating.

📊 heuristic breakdown

Dimension	Signal
Specificity	`███████░░░` 0.68
Control / acceptance	`█████░░░░░` 0.53
Context grounding	`██████░░░░` 0.59
Structure	`█████░░░░░` 0.51
Lexical diversity	`████████░░` 0.85
Clarity (readability band)	`█████████░` 0.86
Length adequacy	`████████░░` 0.79

Readability (display-only): Flesch 67 · FK grade 7.4 · Gunning Fog 10.2

Offline classical-NLP score — no LLM. Higher = more specific, bounded, and grounded prompting.

Tip

✅ Security scan clean

No credentials leaked in this branch's diff. h5i scanned every added line against the secret rule pack and found nothing to rotate.

Note

Duplicate-code scan clean — no copy-paste blocks introduced.

🧠 Reasoning by file — 10134 nodes across 5 lanes, latest 24 only

%%{init: {"flowchart": {"nodeSpacing": 42, "rankSpacing": 48, "diagramPadding": 14}, "themeVariables": {"fontSize": "18px"}} }%%
flowchart TB
  subgraph lane_0["💭 Reasoning · 6 steps"]
    direction LR
    n_lane_0_0["NOTE · DEFERRAL (~/Dev/h5i/docs/index.html): into the previous commit (`9ed22e8…"]:::n
    n_lane_0_1["THINK · Now I'll restructure into six cards, one per command family, with accura…"]:::t
    n_lane_0_2["NOTE · DEFERRAL (~/Dev/h5i/docs/index.html): Committed as a separate follow-up.…"]:::n
    n_lane_0_3["THINK · Let me understand the current structure of MANUAL."]:::t
    n_lane_0_4["THINK · md also has a hand-written `## Table of Contents` (shown on GitHub) I'll…"]:::t
    n_lane_0_5["NOTE · DEFERRAL (~/Dev/h5i/MANUAL.md): r GitHub, where that hand-written TOC ac…"]:::n
    n_lane_0_0 --> n_lane_0_1
    n_lane_0_1 --> n_lane_0_2
    n_lane_0_2 --> n_lane_0_3
    n_lane_0_3 --> n_lane_0_4
    n_lane_0_4 --> n_lane_0_5
  end
  subgraph lane_1["📄 docs/index.html · 8 ops"]
    direction LR
    n_lane_1_0["READ"]:::o
    n_lane_1_1["EDIT × 7"]:::a
    n_lane_1_0 --> n_lane_1_1
  end
  subgraph lane_2["📄 MANUAL.md · 7 ops"]
    direction LR
    n_lane_2_0["READ × 3"]:::o
    n_lane_2_1["EDIT × 2"]:::a
    n_lane_2_2["READ"]:::o
    n_lane_2_3["EDIT"]:::a
    n_lane_2_0 --> n_lane_2_1
    n_lane_2_1 --> n_lane_2_2
    n_lane_2_2 --> n_lane_2_3
  end
  subgraph lane_3["📄 index.html · 2 ops"]
    direction LR
    n_lane_3_0["READ"]:::o
    n_lane_3_1["EDIT"]:::a
    n_lane_3_0 --> n_lane_3_1
  end
  subgraph lane_4["📄 scripts/gen_manual.py · 1 op"]
    direction LR
    n_lane_4_0["READ"]:::o
  end
  classDef o fill:#dbeafe,stroke:#1e3a8a,color:#0b1c4a;
  classDef t fill:#fef3c7,stroke:#92400e,color:#3f2d05;
  classDef a fill:#dcfce7,stroke:#166534,color:#0a2e16;
  classDef n fill:#ede9fe,stroke:#5b21b6,color:#221251;
  classDef m fill:#e5e7eb,stroke:#374151,color:#0b0f17;

Note

🪙 Token reduction — 22 captured tool outputs kept out of context: 17570 → 13671 tokens (22% saved, 3899 tokens). Full output is recoverable with h5i recall object.

📜 Per-commit provenance — 7 AI-authored commits

`81f472a8` docs(manual): rename 'h5i env'/'h5i team' section headings to drop parentheticals

prompt — h5i env (isolated agent sandboxes) and h5i team (auditable agent ensembles) also should be h5i env and h5i team
- 🪴 prompt maturity 27/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`e0df0e2c` docs(manual): rename the 'h5i objects (token reduction)' section to 'h5i objects'

prompt — the section name, h5i objects (token reduction), should be h5i objects
- 🌱 prompt maturity 25/100 nascent — too short, weak context
model claude-opus-4-8 · agent claude-code

`4e5a7fa3` docs(manual): regroup the CLI reference by command family

prompt — Also, let's restructure MANUAL.md (and docs's manula page, which is generated from MANUAL.md). I guess we should group them more precisely based on the sub commands For example. h5i objects is currently one section, but they are part of h5i recall or h5i capture. Also, `h5i…
- 🪴 prompt maturity 42/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`9b9d495d` docs: tighten the six core-capability card descriptions

prompt — I feel each explanation is a bit too long
- 🌱 prompt maturity 14/100 nascent — too short, vague / under-specified
model claude-opus-4-8 · agent claude-code

`d1768eef` docs: map the 'What h5i is' cards to the six command families

prompt — In the front page's "What h5i is An auditable workspace around every agent run.", I guess we explain h5i capture, h5i recall, h5i env, h5i team, h5i audit, and h5i share
- 🪴 prompt maturity 25/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`07ad4c2a` docs: decouple hero embed width from the headline text

prompt — hmm, did you change the format of the h5i serve's example (embed one)? oh, I see. I guess the width of the frontpage is shortened. What is the motivation? Hmmm, it seems like h5i serve's embeded tab's width depends on the width of the title, probably because they live in th…
- 🪴 prompt maturity 27/100 developing — vague / under-specified, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`9ed22e86` docs: lead with Auditable Workspaces category framing

prompt — Do you think we can further improve the speed without any security violation? Can you then review [codex] add injection scanner unit tests #239? Let's further restructure / improve the website in ./doc. "うん、“Auditable Workspace” 寄りにしたほうがいいと思う。今のトップは、eyebrow ではすでに **“AUDITABLE W…
- 🌿 prompt maturity 64/100 proficient
model claude-opus-4-8 · agent claude-code

_{Generated by h5i · re-run h5i share pr post to refresh.}

Koukyosyumei added 9 commits June 29, 2026 22:11

update

475e296

update

3741cd0

docs: tighten the six core-capability card descriptions

9b9d495

Trim each 'What h5i is' card to one or two short sentences.

docs(manual): rename the 'h5i objects (token reduction)' section to '…

e0df0e2

…h5i objects' Drop the parenthetical and update the slug to #h5i-objects across the TOC and every cross-reference; regenerate docs/manual/index.html.

Koukyosyumei changed the title ~~Improve website~~ docs: lead with Auditable Workspaces category framing Jun 30, 2026

Koukyosyumei merged commit c017b21 into main Jun 30, 2026
5 checks passed

Koukyosyumei deleted the improve-website branch June 30, 2026 12:49

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

docs: lead with Auditable Workspaces category framing #244

docs: lead with Auditable Workspaces category framing #244
Koukyosyumei merged 9 commits into
mainfrom
improve-website

Koukyosyumei commented Jun 30, 2026

Uh oh!

Koukyosyumei commented Jun 30, 2026

✅ Security scan clean

`81f472a8` docs(manual): rename 'h5i env'/'h5i team' section headings to drop parentheticals

`e0df0e2c` docs(manual): rename the 'h5i objects (token reduction)' section to 'h5i objects'

`4e5a7fa3` docs(manual): regroup the CLI reference by command family

`9b9d495d` docs: tighten the six core-capability card descriptions

`d1768eef` docs: map the 'What h5i is' cards to the six command families

`07ad4c2a` docs: decouple hero embed width from the headline text

`9ed22e86` docs: lead with Auditable Workspaces category framing

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

Koukyosyumei commented Jun 30, 2026

Uh oh!

Koukyosyumei commented Jun 30, 2026

🪙 78% AI-authored · 4 files · prompt 48/100

💬 The ask

📍 What shipped

✅ Security scan clean

81f472a8 docs(manual): rename 'h5i env'/'h5i team' section headings to drop parentheticals

e0df0e2c docs(manual): rename the 'h5i objects (token reduction)' section to 'h5i objects'

4e5a7fa3 docs(manual): regroup the CLI reference by command family

9b9d495d docs: tighten the six core-capability card descriptions

d1768eef docs: map the 'What h5i is' cards to the six command families

07ad4c2a docs: decouple hero embed width from the headline text

9ed22e86 docs: lead with Auditable Workspaces category framing

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

`81f472a8` docs(manual): rename 'h5i env'/'h5i team' section headings to drop parentheticals

`e0df0e2c` docs(manual): rename the 'h5i objects (token reduction)' section to 'h5i objects'

`4e5a7fa3` docs(manual): regroup the CLI reference by command family

`9b9d495d` docs: tighten the six core-capability card descriptions

`d1768eef` docs: map the 'What h5i is' cards to the six command families

`07ad4c2a` docs: decouple hero embed width from the headline text

`9ed22e86` docs: lead with Auditable Workspaces category framing