Skip to content
Agent Memory
Skill

/storage-manifest

The storage shelf: one card per item archived to object storage (B2, Supabase Storage, S3). Register a card after a verified archive, find a card before any restore, restore through the agent's own restore skill with the card's restore form, then record the restore on the card.

From plugin
gbrain
33k80 skills

Claude Code

$ npx -y skills add garrytan/gbrain --skill storage-manifest --agent claude-code

SKILL.md

storage-manifest.SKILL.md
name: storage-manifest
version: 1.0.0
description: |
  The storage shelf: one card per item archived to object storage (B2,
  Supabase Storage, S3). Register a card after a verified archive, find a
  card before any restore, restore through the agent's own restore skill with
  the card's restore form, then record the restore on the card.
triggers:
  - "bring back"
  - "restore the archived"
  - "where did X go"
  - "what's in B2"
  - "what did we archive"
  - "register this archive"
tools:
  - storage_find
  - storage_get
  - storage_list
mutating: true
writes_pages: false

Storage manifest

> **Convention:** see [conventions/brain-first.md](../conventions/brain-first.md): > look on the shelf with `gbrain storage-find` before listing a bucket or asking the user where something went.

Full guide: `docs/guides/storage-manifest.md`.

Contract

  • **Register after the archive is verified, never before.** Once the upload's checksum or snapshot check passed,

run `gbrain storage-register --card '<json>'` (or pipe the card on stdin) from the trusted local CLI. The card names every copy (store, bucket, path, tool, repository, snapshot, size, hash and its kind, when verified), the original path, why it was archived, and the restore form `{ skill, args }`.

  • **Find before restore.** `gbrain storage-find "<the user's words>" --json` (MCP `storage_find`), then

`gbrain storage-get <id> --json` for the copies and the restore form. A miss lists the closest cards: show them. On `storage_record_invalid` the stored card is broken: tell the user and do not restore from it.

  • **Restore through the agent's own restore skill.** Run the skill the card's `restore.skill` names with

`restore.args`. Cards never carry commands or credentials; the restore skill reads credentials from its own environment. gbrain never contacts a store.

  • **Record the restore.** After the restore, run `gbrain storage-update <id> --expected-revision <rev>

--restored-to <path>` with `--restored-verified` only when the checksums matched. `<rev>` is the revision `storage-get` returned; on `storage_revision_conflict` read the card again and update once.

  • **Interim records** (`stored-object` pages under `storage-manifest/` written before the shelf existed) move with

`gbrain storage adopt`: preview, show the user every refused field, then apply the printed command.

Anti-Patterns

  • Putting a command, a password, a signed URL or an environment assignment on a card: it is refused.
  • Writing the shelf with `put_page` or `remember`: refused `storage_source_managed`.
  • Restoring from a bucket listing without reading the card first.
  • Recording `--restored-verified` without checking the hashes.

Output Format

Tell the user what the card says in plain words: what the item is, where each copy lives, its size, when it was last verified, and which skill restores it. After a restore, say where it landed and whether the checksums matched.

When it fails

Read the refusal's `code` and follow its `fix` (see [the agent operator protocol](../conventions/agent-operator-protocol.md)). `storage_manifest_missing` on a brain with a Git checkout means the shelf is not hosted yet: relay the `gbrain storage init` preview to the user.

Read more
@garrytan
Ships withgbrain

Garry's Opinionated OpenClaw/Hermes Agent Brain

Stars on GitHub

Momentum

stars in total
33.0k
Star history of garrytan/gbrain on GitHub, a chart by star-history.com

Other skills on gbrain.

brain-ingest-gate

brain-ingest-gate

Pre-write quality gate for content entering the brain. No raw copies: a bare cp/mv into the brain repo is a bug. Before any new…

@garrytan@garrytanInstall Skill
brain-link-discipline

brain-link-discipline

When you report a brain page to the user — created, edited, committed, or relayed from a subagent — a working link is part of the…

@garrytan@garrytanInstall Skill
brain-ops

brain-ops

Brain knowledge base operations. The core read/write cycle: brain-first lookup, read-enrich-write loop, source attribution,…

@garrytan@garrytanInstall Skill
concept-synthesis

concept-synthesis

Deduplicate and synthesize raw concept stubs into a tiered intellectual map (T1 Canon to T4 Riff), tracing idea evolution across…

@garrytan@garrytanInstall Skill
context-audit

context-audit

Token-hygiene audit of the always-loaded context stack — CLAUDE.md, AGENTS.md, auto-memory MEMORY.md, and the bootstrap-rendered…

@garrytan@garrytanInstall Skill
correction-pipeline

correction-pipeline

When the user corrects a factual error, root-cause it immediately. Don't just note the correction — trace the error to its…

@garrytan@garrytanInstall Skill

More skills from other plugins.

agent-cost-report

agent-cost-report

Believable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude…

@thedotmack@thedotmackInstall Skill
bot-handoff

bot-handoff

Keep a durable record of work passed between Bots in the shared Hindsight memory bank. Use when handing work to another Bot, when…

@vectorize-io@vectorize-ioInstall Skill
mempalace

mempalace

Install, configure, and operate MemPalace, including a private local palace, a shared-brain hub, or a client joining an existing…

@mempalace@mempalaceInstall Skill
mempalace

mempalace

Install, configure, and operate MemPalace, including a private local palace, a shared-brain hub, or a client joining an existing…

@milla-jovovich@milla-jovovichInstall Skill

Questions about storage-manifest.

Which coding agents can use storage-manifest?
storage-manifest installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.