Social Ledger Convention

Research date: 2026-06-26. Canonical package asset: assets/social-ledger-convention.md. Parent router: docs/social-post-convention.md.

This is the system-of-record contract for alignment-producing agents that draft, approve, post, and promote social content. It defines how an agent finds prior posts, avoids duplicates, records what was drafted and published, links public-safe context, and keeps canonical records private. Load this file when a social drafting workflow needs cross-run memory of social activity or a public alignment context link; it is not needed to draft a single one-off post with no ledger.

Resolution Order

Resolve the ledger by posting identity first, then project. A single account can post about multiple projects, so duplicate detection and promo cadence are account-scoped, not only repo-scoped. A project that owns its own account can keep a project-local ledger with no central account ledger.

Resolve scope from project config before drafting, not by asking on every run. Read these fields from .agents/project.json under a social_ledger key (or a project-documented equivalent):

Field	Meaning	Example
`mode`	`local`, `central`, or `both`	`central`
`account_id`	Posting identity that owns the ledger	`george-x`
`central_repo`	Private repo holding central account ledgers	`GeorgeQLe/me`
`central_path`	Path template inside the central repo	`social-ledger/accounts/<account-id>/posts`
`local_path`	Path for project-local records	`social/posts`

If config is missing, default to the local project ledger unless the user explicitly selects a central account ledger, and offer to persist the resolved scope to project config so later runs do not re-ask.

Ledger Scope

Scope	Use When	Canonical Path	Agent Behavior
Project-local ledger	The project owns the posting account, or posts do not need cross-project dedupe.	`social/posts/*.yaml` in the project repo.	Read and write only local records for that project/account.
Central account ledger	One personal or brand account posts across many projects.	`social-ledger/accounts/<account-id>/posts/*.yaml` in the configured private central repo, defaulting to `GeorgeQLe/me`.	Read account-wide history before drafting; write approved drafts and posted URLs to the central account ledger.
Both local and central	The project needs local provenance and the shared account needs account-wide dedupe.	Local mirror plus central canonical account record.	Prefer central for dedupe and status; keep a local pointer or copy for project provenance.
Public gBrain projection	A public-safe alignment doc should provide context for a post.	`https://6eorge.com/brain/<project-id>/<doc>.html`.	Link only approved public-safe docs; never publish private ledger records.

The canonical ledger is private by default. The only public surface is the sanitized alignment archive link and any post metadata the user explicitly approves.

Record Shape

Each record must be stable enough for a later agent to answer what has already been drafted, approved, posted, linked, and promoted.

id: 2026-06-25-social-ledger-public-archive
account_id: george-x
project_id: agentic-skills
channel: x
topic: Social ledger and public archive
status: approved_draft
source_alignment_page: alignment/social-ledger-public-archive.html
public_alignment_url: https://6eorge.com/brain/agentic-skills/social-ledger-public-archive.html
drafting_mode: creator_inspired
post_mode: post_plus_replies
main_post: "..."
reply_chain:
  - purpose: full_alignment_doc
    text: "Full alignment doc for context: ..."
    url: https://6eorge.com/brain/agentic-skills/social-ledger-public-archive.html
  - purpose: skill_promo
    text: "The workflow is powered by skillpacks: ..."
    url: https://www.npmjs.com/package/skillpacks
published_urls: []
dedupe_fingerprint: "x:agentic-skills:social-ledger-public-archive"
risk_notes: []

Statuses: draft, approved_draft, posted, skipped, needs_revision, retired. published_urls stays empty until a human posts manually and records the live URL.

reply_chain purposes: full_alignment_doc (public gBrain context link), skill_promo (contextual install/source link), or another labeled purpose. Keep one purpose per reply.

Duplicate Detection

Before drafting, read the resolved ledger (account-wide for central/both, project-local for local) and check the dedupe_fingerprint of prior records.

Fingerprint shape: <channel>:<project_id>:<topic-slug>.
If a matching fingerprint already has a posted record, do not re-draft the same angle silently. Surface the prior post (and its published_urls) and ask whether to skip, post a genuinely new angle, or follow up.
Promo cadence is account-scoped: avoid repeating the same skill-promo wording across recent posts on the same account even when projects differ.

Public Archive Link Handling

Public-safe alignment context links resolve to the gBrain archive at https://6eorge.com/brain/<project-id>/<doc>.html, served from the private GeorgeQLe/me repo's /brain route. The agentic-skills source alignment HTML is synced into gBrain by the configured alignment-archive sync.

Before proposing a public_alignment_url, the agent must verify the source page is safe for public viewing and is either already synced to gBrain or approved to sync. Do not invent a public URL for a page that has not been published.

Public Safety Boundary

Keep canonical ledger records private. Publish only public-safe alignment docs and post metadata the user explicitly approves — users approve what is put out publicly.

Do not publish: secrets, credentials, internal account tokens, private repo details, confidential customer names, private research inputs, unverifiable metrics, or recurring reply templates that create spam risk. This boundary is in addition to, not a replacement for, the source-safety rules in docs/social-post-convention.md.

X Posting Pattern

The preferred X output is post_plus_replies: a main insight post, a first reply carrying the public alignment context link, and an optional second reply with a relevant skill/package link. See docs/social/x-post-convention.md for the channel-level shape; this convention governs how those parts are recorded in the ledger.

Main post must stand alone as the insight, lesson, result, or decision. It must not depend on the link to be useful.
First reply uses a public gBrain URL only when the alignment doc is approved for public context.
Optional skill reply is used only when the post is actually about Agentic Skills, the workflow, or a reusable skill. Rotate wording to avoid repetitive promo.

Contextual skill-promo replies are allowed with rotated wording. Default the promo target by intent:

npm (https://www.npmjs.com/package/skillpacks) when install or use is the point.
GitHub when source, docs, or contribution context matters.

Do not generate a promo reply when the post is not genuinely about the skills/workflow, and do not reuse identical promo wording across recent posts on the same account.

Out Of Scope (This Build)

Automation scripts (ledger YAML validation, account-history summary, central sync/read helpers) are deferred. This convention defines the contract; add scripts only under separate approval.