The hub's install guide told Codex and Hermes users to run four CLI
commands by hand, and the one people skipped — `bdrive hooks install` —
is exactly the one that makes files sync at turn boundaries. Hand the
setup to the agent instead, the way the Claude tab hands it to the
plugin.
- `bdrive skill install` (internal/agentskills, plugin/embed.go): the
binary now carries the beardrive skill and writes it to any agent that
reads SKILL.md — ~/.{claude,codex,gemini,hermes}/skills/beardrive/.
User-level on purpose: the skill is about the CLI, not one folder, and
a synced project folder should never carry it. Idempotent; refreshes a
stale copy after a CLI upgrade. Bare `bdrive skill` prints the table,
mirroring `bdrive hooks`.
- Guide's Codex/Hermes tabs are now a single paste, no terminal: the
prompt has the agent install the CLI, keep the skill, sign in, init,
and register hooks. The commands ride inside the prompt because these
agents ship no BearDrive knowledge (Claude's tab is terse only because
the plugin carries it). `login --device` there — a browser-callback
sign-in is invisible to an agent mid-turn, while the device flow gives
it a code and URL to relay. Plain commands live on in an "or run it
yourself" fallback.
- Docs realigned: README, SKILL.md, /beardrive:install, self-hosting.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
From the first agent-conversation e2e against a live hub:
- install.md: git handoff for tracked knowledge folders (one transport per
folder), .bdrive//.bdriveignore git guidance, Hermes per-user hook warning,
drop stale share--list aside
- SKILL.md: .bdriveignore is local-only on --shared mounts (root file sits
outside the include list) — doc now matches behavior
- new project skill .claude/skills/onboarding-e2e: reusable procedure to
re-run the conversation-level e2e (staging, instruction materialization,
agent prompt contract, verification, pass bar)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VbiaaVM2ACxeRi8ySG9ybc
- login copy across README/skill/commands: bare `bdrive login` (BearDrive
Cloud) is the default; signup auto-creates a free personal workspace,
pending invites route into the team; self-hosted hubs pass a URL
- init flows propose a detected knowledge folder (wiki/docs/notes/handbook/
vault) as --shared instead of asking open-endedly; repo-root rule restated
- install always ends by handing the user a `bdrive url` link to a
representative page — the activation moment
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VbiaaVM2ACxeRi8ySG9ybc
Field report follow-up: an agent with a stale skill copy couldn't find
the gated URL after creating a file. Instructions rot; hook output is
computed fresh from the binary every turn. Claude Code's turn-start
pull hook now runs 'bdrive sync --hook claude-code', which:
- pulls as before, and stamps the session note from the event JSON
(replacing the sh/sed pipeline for the pull leg)
- emits the project's gated-link formula as UserPromptSubmit
additionalContext: whenever the agent mentions a synced file path in
prose, it appends the hub link on an emoji — `<path>` [🔗](<url>) —
path plain (it's the local path), hyperlink on the emoji only; code
blocks stay plain; bdrive share stays explicit-opt-in-public
Blind-tested: an agent given only the injected context decorated every
path mention correctly, kept the code-block command plain, and checked
files were synced before linking.
- hooks install now CONVERGES marker-identified groups to the current
shape (command/matcher/flags), so improvements reach existing
projects on reinstall instead of being frozen by the idempotency
marker; hermes same
- plugin: UserPromptSubmit → beardrive-pull.sh (stdout passes through);
version 0.3.0
- SKILL 'Share what you make' generalized to 'Link what you mention'
(URL formula documented for non-Claude platforms); install.md pointer
template updated
Never fails the turn: every error path in --hook mode is a silent
successful exit; offline still emits (links serve online teammates).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
Reviewer round 4 (9/6/9/9/7/9) caught that round 3 only fixed the
Quick start: the bare-login default survived in the four highest-
traffic surfaces. All inverted to self-host-first:
- README HERO code block (+ the other-machine one-liner and the
every-other-device shorthand): bdrive login https://your-hub, with
Cloud as a clearly-labeled waitlist callout
- plugin/commands/install.md + init.md step 2: ask for the team's hub
URL; bare 'bdrive login' explicitly flagged as targeting the
not-yet-open Cloud
- SKILL.md walkthrough step 1 + the hub-example comment
- website hero terminal animation: 'syncing → your team's hub', share
URL on a neutral acme-hub.example host (og:url and the coming-soon
card untouched — those are honest)
- README's login reference sentence now says the beardrive.ai default
is waitlist-only
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
New command printing a file's hub viewer URL: sign-in + project-org
membership required to open (the internal counterpart to bdrive share's
public links). Computed locally from the mount config — hub origin +
project id from the remote, path segments percent-encoded with literal
slashes; unsynced paths (ignored, or outside a --shared scope) are
refused so nobody gets handed a 404. --sync pushes first so a
just-created file's link resolves immediately; no arg = project home.
The plugin docs now instruct agents to include this link in their reply
whenever they create a shareable artifact (.md/.html/.csv/...) in the
shared folder, reserving bdrive share for people outside the hub:
SKILL.md command map + 'Share what you make' guidance, install.md root
pointer template + payoff step, README, CLAUDE.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
Field report: an agent session read a dozen-plus files via Bash
(grep/cat/tail/find) and Grep, but read heat showed almost nothing —
the read-log hook only matched the native Read tool.
- matchers broadened per platform: claude Read|Grep|Bash, codex
read_file|shell, gemini +search_file_content|run_shell_command,
hermes read_file|grep|bash; plugin hooks.json matches Read|Grep|Bash
- read-log is now tool-aware: shell events mine the command line for
existing files it names (redirect targets and flags excluded), grep
events mine the response for the files the matches came from
(content lines and filename lists), and listing tools (Glob, ls) are
deliberately ignored — seeing a file's name is not reading it
- hooks install upgrades a registered hook's stale matcher in place, so
re-running it after a binary upgrade rolls coverage out to existing
projects instead of being skipped by the idempotency marker
- docs: SKILL.md platform table + read-heat wording, install/init
commands, README command table
Note from the same report, verified not a bug: read-log resolves the
mount via the folder's own .bdrive/config.json (config.ResolveMount),
so a stale duplicate registry entry cannot swallow reads.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
Replace the old "append conventions to CLAUDE.md" step with a two-file
pattern for teaching agents a newly mounted shared folder:
- <shared>/AGENTS.md (synced): the team's single source of truth for the
folder's structure and conventions — scaffolded once by the project
creator, read (never rewritten) by joiners.
- A repo-root AGENTS.md/CLAUDE.md pointer (per machine, not synced): the
awareness/routing layer. Required because discovery differs by
platform — Claude Code and Hermes load nested instruction files lazily
(only after entering the folder), and Codex never discovers them at
all (root→cwd path only).
SKILL.md gains a "Teaching agents the folder" section with the platform
discovery table and a first-contact orientation ritual; /beardrive:install
step 4 and /beardrive:init step 6 now offer both files as separate
consents; install.md step 3 gains the same never-sync-a-repo-root hard
rule init.md already had; README/CLAUDE.md descriptions updated.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
Admin/org-owner Insights view (⋯ menu): dependency-free SVG scatter of
every file by 30-day reads × days since last change, log scales, with the
hot-but-stale danger quadrant shaded and a ranked fix-these-first list;
lens toggle for all/human/agent reads. The Claude Code plugin gains a
PostToolUse(Read) hook so plugin users feed agent-read telemetry without
project-level hook registration. Docs synced: README, SKILL.md,
plugin install/init commands, CLAUDE.md, design doc marked implemented.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
`bdrive hooks install` detects the agent platforms in use — Claude Code
(.claude/), Codex (.codex/), Gemini CLI (.gemini/), Hermes (~/.hermes/) —
and idempotently merges beardrive's turn-boundary sync hooks into each
platform's own hook config (JSON for claude/codex/gemini, YAML for
hermes), preserving existing hooks. All four pipe hook JSON with a
session_id, so one POSIX-sh hook command serves every platform: pull at
turn start, push after edits, changes stamped "<agent> session <id>".
Bare `bdrive hooks` prints the detection/registration table.
The beardrive skill now runs it automatically after `bdrive init`, and
/beardrive:install's hand-maintained settings.json block is replaced by
the command, so the hook content has one source of truth in the binary.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P5cxPQdSGJnjXCYY9GeWXt
Share links (bdrive share <file>, or the web UI's Share button):
- Mint an unguessable public URL (/s/<token>) anyone can open — no
account. HTML renders as a page, markdown gets a standalone shell,
PDFs open inline; ?download=1 attaches.
- Sandboxed: /s/* responses carry CSP `sandbox allow-scripts` + nosniff
and never see auth cookies, so shared content's scripts run in an
opaque origin and can't touch hub sessions.
- Links serve the file's LATEST synced content and live until revoked;
--expires makes self-destructing ones; --list/--revoke manage them.
Re-sharing a file returns the same link. File-backed shares.json.
- CLI resolves the project by walking up to .bdrive/ from the shared
file, warns when the hub address is private (LAN-only links), and
hints when the file hasn't synced yet.
/beardrive:install (plugin command) — team onboarding driven by Claude:
- Ensures the bdrive binary, signs in (bdrive login), runs bdrive init
(whole folder or a shared subfolder like wiki/).
- Asks before appending a CLAUDE.md section that teaches agents to put
shareable artifacts in the shared folder and mint URLs with bdrive
share; asks before registering project-level hooks in
.claude/settings.json: blocking pull at UserPromptSubmit, async push
on PostToolUse Write|Edit — teammates sync with or without the plugin.
- Fix: the plugin hook script still checked for the old `.bdrive` file
and was a silent no-op since the directory change; now checks -d.
Tests: share creation gating, public access + sandbox headers, dedupe,
latest-content semantics, revoke, expiry, markdown/download variants,
list filtering, registry persistence. Docs updated (README sharing +
Claude Code sections, SKILL.md, CLAUDE.md).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01R7Q9ZKSZRTdvrSJkYLUmYs
