# OaC File Format v1: the machine-readable vocabulary. # # The human standard is specs/oac-file-format.md. The decision is ADR-098. # This file is what `code/scripts/oac-validate.py` reads. One file, so there is # exactly one place a value can be legal. # # THE RULE THAT KEEPS THIS SHORT: # `type` says what KIND of document this is, never what it is ABOUT. # Subject goes in `topics`. Provenance goes in `tags`. # Without that rule this list grows back. It reached 198 values before v1. version: "1.13" # quoted: unquoted, YAML reads 1.10 as the number 1.1 # ── How values are written ─────────────────────────────────────────────────── # Lists MUST use flow style: `topics: [positioning, pricing]`. # NOT block style: # topics: # - positioning # # This is not taste. The shipped reference client parses frontmatter with a # scalar-only splitter (products/oac-client/src/lib/parseSelfAsCode.ts): # "arrays/nested blocks are ignored" / "skip list items / nested / blank" # Under block style it sees `topics` as an empty string and the values vanish. # Under flow style it sees the raw string `[positioning, pricing]`, which it can # store, round-trip and later parse. A format its own reference implementation # cannot read is fiction, so the format bends here rather than the client. array_style: flow # ── Required on every OaC document ─────────────────────────────────────────── required: - title - type - audience recommended: - summary # one plain sentence; feeds generated index.md - status - lifecycle - owner - updated - topics - tags # Field names that mean the same thing as a canonical one. Read by consumers # and by the backfill; the left side is not legal to write in new documents. # `last_updated` is the big one: ADR-094 mandated it and 172 pack files carry # it. `related_adrs` came from STRUCTURING-GUIDE and is just a typed edge. field_aliases: last_updated: updated # READ as `references`, but do not mechanically rename it: its values are # bare ADR numbers, and `references` values must be paths that resolve. # Converting them is ambiguous, because SC-era and OpsCode ADR numbers # overlap. 43 files, awaiting a human. related_adrs: references # ── type: what KIND of document ────────────────────────────────────────────── # 26 values. Each has a real home in the tree. If a document does not fit, # it goes to the manual bucket and a human decides. There is deliberately no # `document` fallback: a catch-all is how 198 values happened. types: adr: Architecture decision record spec: Technical specification or design standard: An org-wide rule we hold ourselves to template: A thing to copy and fill in index: Directory listing or map, for progressive disclosure lesson: Curriculum material guide: How to do a thing, for a human strategy: Direction, positioning, plans, identity market-intel: What the market and competitors are doing audience-def: "Who we sell to: personas, ICPs, buyer language" messaging: What we say to them offer: What we sell and for how much brand: Voice, visual identity, brand assets engagement: A customer or partner relationship document session: Notes or summary from a live session coach-feedback: Coach-facing assessment of a session transcript: A verbatim record of what was said correspondence: Email and messages sent or received contract: Commercial agreements and money documents campaign: A marketing or outreach campaign build-log: A documented build journey (Rule 8) report: Point-in-time analysis, audit or status reference: External system documentation we did not write case-study: A story we tell about work we did agent: An agent definition or prompt profile: A person # ── The 198-to-26 mapping, used by the backfill and kept for provenance ────── # Left: every value found in the repo on 2026-08-08. Right: canonical type. # Keep this table. It is the audit trail for the collapse, and the backfill # reads it rather than guessing. type_aliases: decision: adr design-spec: spec signature-spec: spec design: spec process-standard: standard style-guide: standard sop: standard methodology: standard deployment-template: template email-template: template course-template: template scaffold: template overview: index map_overview: index navigation: index maps_index: index readme: index context_core_readme: index context-core-overview: index context-core: index team-hub: index sales-playbook-index: index campaign-index: index power_capsules_index: index shared_index: index curriculum: lesson curriculum-module: lesson facilitator-script: lesson internal-guide: guide installation-guide: guide installation: guide handover: guide working-doc: guide prompt: guide prompt_reference: guide deployment: guide strategic_identity: strategy strategic-identity: strategy go_to_market: strategy gtm: strategy gtm-advisory: strategy marketing-strategy: strategy cohort-design: strategy launch-plan: strategy living_plan: strategy living-plan: strategy shared_strategy: strategy shared_system: strategy shared_framework: strategy shared_founder: strategy self_as_code: strategy organisation: strategy project: strategy milestone: strategy market_intelligence: market-intel market-intelligence: market-intel market: market-intel competitive_intelligence: market-intel competitive_intel: market-intel competitive-intelligence: market-intel competitor-analysis: market-intel battle-card: market-intel targeting-matrix: market-intel company_lists: market-intel companies-partners: market-intel companies_partners: market-intel persona: audience-def icp: audience-def audience-definition: audience-def audience_definition: audience-def customer_intelligence: audience-def customer-intelligence: audience-def echo-persona-profile: audience-def buyer-pains: audience-def buying-triggers: audience-def objection-library: audience-def customer-vocabulary: audience-def messaging_framework: messaging messaging_frameworks: messaging email-sequences: messaging linkedin-sequences: messaging campaign-messaging: messaging booking-page-copy: messaging sales-playbook: messaging # A customer's revenue playbook is sales enablement content, same as a # sales playbook. 94 files across two engagements carried this folder name # and no type, which is what left them below level 1 after the first pass. revenue-playbook: messaging # Hybrid: it is a messaging document; the persona is the subject, so it # belongs in `topics`, not in `type`. Exactly the rule this list rests on. persona_messaging_strategy: messaging value_stack: offer value-stack: offer value-offers: offer value_offers: offer lead-magnet: offer cohort proposal: offer pre-engagement proposal: offer branding: brand brand_manifest: brand brand-asset: brand brand_asset: brand personal_brand: brand voice: brand voice_stream: brand voice-stream: brand voice-source: brand shared_voice: brand asset: brand assets: brand # The SPEAK family collapses to brand plus a speak_component facet. speak_framework: brand speak_framework_core: brand speak_framework_implementation: brand speak_implementation: brand speak_story: brand speak-story: brand speak_purpose: brand speak-purpose: brand speak_educate: brand speak-educate: brand speak_authority: brand speak_keep_consistent: brand speak-keep-consistent: brand speak-framework: brand client: engagement engagement_analysis: engagement customers: audience-def engagement-artefact: engagement session-summary: session session-ai-summary: session coach-rollup: session coaching-note: session meeting-prep: session session_transcript: transcript session-transcript: transcript engagement_correspondence: correspondence engagement_email: correspondence engagement-email: correspondence client-email: correspondence email-sent: correspondence email-samples: correspondence comms: correspondence comms-artefact: correspondence comms-draft: correspondence discovery-followup: correspondence attachment: correspondence commercial-agreement: contract agreement_draft: contract purchase-request: contract invoice-line-item: contract cost-estimate: contract campaign-strategy: campaign campaign-plan: campaign campaign-overview: campaign campaign-offer-brief: campaign audit: report review: report analysis: report reports: report pipeline-status: report growth_modeling: report growth-modeling: report lead-gen-cost-forecast: report key-result-areas: report key_result_areas: report daily-log: report task-bridge: report team-task-bridge: report changelog: report enrichment: report extraction-manifest: report planning-artifact: report research-note: report pre-engagement it compliance review: report api_documentation: reference crm-configuration: reference sample-account: reference sample-brief: reference opscode_resource: reference article: case-study content-idea: case-study agent-lab-pointer: agent team-profile: profile role: guide # ── The self-plane collapse (added for ADR-118) ──────────────────────────── # A self plane holds a life, not just a business, so it carries document # kinds an operations repo never sees: fitness programs, reading lists, # cleaning rotas, family planning. The SAME rule still settles every one of # them, and that is the point worth teaching: `type` is what a document IS. # # A fitness program, a cleaning rota and a revenue target are all # 'an ongoing rule you hold yourself to' -> standard # A fitness log, an expense tracker and a retro are all # 'where things stood when I looked' -> report # A vision, an OKR, a weekly plan and a venture roadmap are all # 'where I am trying to get to' -> strategy # # 26 types held. Nothing here needed a 27th, which is the strongest evidence # yet that the collapse from 198 was real rather than lucky. # `strategic-identity` and `project` already map to strategy above; the # self plane simply proved them right a second time. vision: strategy annual-goals: strategy quarterly-okr: strategy monthly-targets: strategy weekly-plan: strategy venture-goals: strategy venture-milestones: strategy product-roadmap: strategy curriculum-plan: strategy notable-plan: strategy offer-framework: offer engagement-playbook: guide ip-positioning-brief: strategy ip-ownership-brief: contract wind-down-runbook: guide shared-planning: strategy travel-plan: strategy fitness-program: standard cleaning-schedule: standard home-maintenance: standard recurring-tasks: standard monthly-budget: standard revenue-targets: standard hiring-plan: strategy role-definitions: standard fitness-log: report quarterly-retro: report expense-tracker: report purchase-tracker: report skills-tracker: report investment-notes: report risk-register: report ops-snapshot: report reading-list: reference learning-path: reference setup-guide: guide role-pointer: profile performance-notes: profile venture-advisors: profile venture-decisions: index voice-stream-overview: index # A rolling task list is 'where things stood when I looked', the same kind of # document as a status report. The precedent is this repo's own # team/task-bridge.md, which has carried `type: report` since January. inbox: report email-actions: report action-items: report dashboard: report meeting-transcript: transcript process: guide # An owner's private read on a partner. Not `engagement`: that is the # relationship document itself, which lives in the shared repo. This is the # analysis of it at a moment in time, and it never leaves the self plane # (ADR-111). private_strategic_read: report # Values with no safe mapping. The backfill must NOT guess these; it lists # them for a human. `internal` is a visibility statement wearing a type's # clothes, which is exactly the confusion `audience:` exists to end. type_manual_review: - internal # Added 1.5. These three describe a SUBJECT AREA, not a kind of document, so a # single target is always wrong for some of the files carrying them. In one # cohort pack `business_knowledge` covered a price list, a fabrication rule, a # materials reference and a positioning statement: `offer`, `guide`, `reference` # and `strategy` respectively. Until 1.5 all three aliased to `engagement`, # which means a customer or partner relationship document and was wrong for # every one of them. - business - business_knowledge - operations # ── Where a document lives implies what KIND it is ─────────────────────────── # The other half of the backfill, and the half that was missing from the # standard until 2026-09-12. It lived as a hardcoded table inside # code/scripts/backfill-frontmatter.py, which meant the rule we ask every # operator to follow was one we could not hand them. Now it ships. # # ORDERED: first match wins, so the specific precedes the general. # NO FALLBACK: a path that matches nothing is left for a human. A fallback is # how a vocabulary re-inflates, and this one reached 198 values before v1. # # Keyed by plane (ADR-118), because the two planes have different furniture. # A business plane has engagements, campaigns and curriculum. A self plane has # a spine, five folders filed by actionability (SPADE, ADR-126), and a life in # them. path_types: business: - ['(^|/)(README|index|overview)\.md$', index] - ['(^|/)CLAUDE\.md$', guide] - ['^SECURITY\.md$', standard] - ['facilitator-script', lesson] - ['^decisions/\d+-', adr] - ['^specs/', spec] - ['^brand/standards/', standard] - ['^build-logs/', build-log] - ['^reference/', reference] - ['^templates/', template] - ['^team/', profile] - ['-standard\.md$', standard] - ['^curriculum/', lesson] - ['-coach-feedback\.md$', coach-feedback] - ['coach-rollup\.md$', coach-feedback] - ['(^|/)transcripts?/', transcript] - ['-transcript\.md$', transcript] - ['-session-notes\.md$', session] - ['(^|/)sessions?/', session] - ['(^|/)campaigns/.*(email|linkedin)-', messaging] - ['(^|/)(email|comms)-[^/]*\.md$', correspondence] - ['proposal', offer] - ['prospect-brief\.md$', engagement] - ['engagement-(memo|shape|plan)\.md$', engagement] - ['cohort-plan\.md$', engagement] - ['task-bridge\.md$', report] - ['^campaigns/|(^|/)campaigns/', campaign] - ['^(icp|marketing|sales|gtm|services|content)/', strategy] - ['(^|/)reports?/', report] - ['(^|/)getting-started/', guide] - ['(^|/)research/', market-intel] - ['(^|/)(drive-ingest|documents)/', reference] - ['(^|/)100-day-plan/', strategy] - ['raise-the-floor', template] - ['(battlecard|battle-card|-comparison)\.md$', market-intel] - ['(architecture|-brief|-schema|-mockup|-design)\.md$', spec] - ['(curriculum-|per-role-prework)', lesson] - ['week-\d+-tasks\.md$', engagement] - ['(pricing|cohort-sizing)', offer] - ['(-configuration|config|tracking-links)\.md$', reference] - ['(welcome-message|^invite-|/invite-)', correspondence] - ['meeting-summary', session] - ['(marketing-plan|mutual-goals)', strategy] - ['-strategy\.md$', strategy] - ['(^|/)plans?/', strategy] - ['(^|/)drafts/', correspondence] - ['(^|/)[^/]*[Ee]vent[^/]*/', engagement] - ['^STRUCTURING-GUIDE\.md$', standard] - ['speak-framework', brand] - ['handover', guide] - ['(rollup|actions)\.md$', report] - ['^products/', spec] - ['^infrastructure/', spec] - ['^brand/personal-branding/', brand] - ['^agents/', agent] - ['^\.github/', template] - ['^brand/', brand] # ── The self plane ───────────────────────────────────────────────────────── # Derived from the first real migration (lopez-delon-ops, 2026-09-12). Every # rule here generalises: the next operator's `standards/` also holds bars # they hold, and their `archive/` is also done or dropped. # From 1.11 the folders are SPADE (ADR-126): `archive/` was `history/` and # `evidence/` was `library/`. The old names stay below as legacy rules, # because a repo made before 1.11 keeps its folders until its owner renames # them. self: - ['(^|/)(README|index|overview)\.md$', index] - ['(^|/)CLAUDE\.md$', guide] # The spine. Its own profile (`self-core`) describes the body grammar; # this says what kind of document it is. - ['^self-as-code\.md$', strategy] - ['^decisions/\d+-', adr] # The thing you copy to make the next ADR is a template, not an ADR. # Must follow the numbered rule and precede nothing in particular. - ['(^|/)template\.md$', template] # A guide is a guide wherever it sits. Must precede the templates rule, # which would otherwise call a setup walkthrough a template. - ['-(guide|setup|reference)\.md$', guide] - ['^evidence/templates/', template] - ['^evidence/', reference] - ['^library/templates/', template] # legacy name for evidence/ - ['^library/', reference] # legacy name for evidence/ # The SPADE folders. `standards/` is the bar you hold, so a document in it # is usually the statement of that bar. - ['^standards/self/(goals|weeks)/', strategy] - ['^standards/self/vision\.md$', strategy] - ['^standards/ventures/[^/]+/(goals|milestones)', strategy] - ['^standards/ventures/.*/partnerships/', report] - ['^standards/(household|finance)/', standard] - ['^projects/', strategy] # `archive/` says the document is finished, never what kind it is, so it # is classified by what it was. - ['^archive/plans/', strategy] - ['^archive/team/', profile] - ['^archive/ventures/', strategy] - ['^history/plans/', strategy] # legacy name for archive/ - ['^history/team/', profile] # legacy name for archive/ - ['^history/ventures/', strategy] # legacy name for archive/ # Shared surfaces the post-session chain and the sweep produce. - ['(^|/)sessions?/', session] - ['-transcript\.md$', transcript] - ['-private-watch\.md$', report] - ['(^|/)correspondence/', correspondence] - ['(^|/)meetings/', session] - ['^inbox\.md$', report] # The audience a backfill stamps on a document that carries none. Keyed by # plane, and FAIL CLOSED in both: never stamp a value more shared than the # plane's own floor. A self plane is owner-only by definition (ADR-111), so # `owner` is both its floor and its constant. A business plane's floor is # `internal`, never `team` or `public`, because over-sharing is the one # mistake a backfill cannot undo. audience_default: business: internal self: owner # ── audience: who may read it ──────────────────────────────────────────────── # Load-bearing. code/scripts/audit-sensitive-content.py greps this field, and # the gate is blind to any file that omits it. audience: owner: The owner only. In a shared repo, move it to the owner's own repo. internal: The organisation, including Tier 1 delivery partners. team: Named collaborators on a specific engagement. public: Safe to publish. # `audience_note` is an OPTIONAL free-text companion, for when the real # readership is richer than four values can hold: 'internal + share-ready with # Luke', 'Operators and founders with zero dev background'. `audience` stays # the enum a tool routes on; the note is for people. It exists because 22 # documents had written that sentence into `audience` itself, and collapsing # them to the enum would have thrown the sentence away. # `audience` is a ROUTING signal, not a security boundary. It says where a # document may be shown. It enforces nothing; the git host does that. Tools # that gate on it MUST fail closed: no `audience` means most restricted. # `adam` was the drafting spelling of `owner`, renamed for two reasons. # A published open standard cannot carry one person's name as an enum value: # no other operation has an 'adam'. And the value collided with real people # called Adam, which is why audit-sensitive-content.py had to stop prefix- # matching after it flagged a client-facing proposal for a client named Adam. # The gate reads both spellings until the backfill collapses them. audience_value_aliases: adam: owner # Legacy field spellings the gate reads until the backfill collapses them. # 171 files across three field names before v1. audience_aliases: internal_only: audience # boolean; true -> internal visibility: audience # same value set # ── lifecycle: how the document behaves over time ──────────────────────────── # The consequence that matters: only `record` may be quoted verbatim by an # agent. Quoting a `living` document as speech is how a coach's summary ends up # attributed to a partner. lifecycle: static: A reference the operation reads and does not change. living: Updated as the operation runs. record: Written once, never edited. The only class safe to quote verbatim. lifecycle_by_type: static: [standard, spec, template, reference, brand] living: [strategy, messaging, lesson, engagement, index, campaign, offer, market-intel, audience-def, guide, agent, profile] record: [transcript, session, coach-feedback, correspondence, build-log, contract, report, case-study, adr] # ── status ─────────────────────────────────────────────────────────────────── # Ours, not OKF's. Their spec defaults to `stable`; we have richer states in # use across 834 files and the ADR series depends on them. status: common: [draft, active, complete, superseded] adr: [proposed, accepted, rejected, superseded] pack: [seeded, draft, confirmed] # ADR-094 lifecycle, kept deliberately # ── Fields some types add ──────────────────────────────────────────────────── # The three required fields apply everywhere. These are the EXTRA fields that # carry weight for particular kinds of document. All optional: a missing one # never costs a conformance level. # # This absorbs the five per-context schemas that used to live in # STRUCTURING-GUIDE.md, which was one of the four rival definitions ADR-098 # set out to end. Those schemas invented `type` values that do not exist here # (memo, plan, proposal, brief, summary, timeline). The kind of document is # now the `type`; what it is about is `topics`; and the fields below are the # handful of genuinely per-type extras. type_fields: adr: [adr, date, decision_date, supersedes, superseded_by, formerly] spec: [version] standard: [version] lesson: [module] engagement: [engagement, date] correspondence: [engagement, date] contract: [engagement, date] offer: [engagement, date] session: [engagement, session_date, source] coach-feedback: [engagement, session_date] transcript: [engagement, session_date, speaker_attribution, speakers, source] build-log: [date] # ── Edges: flat keys, never a nested block ─────────────────────────────────── # Flat because `grep -rn '^depends_on:'` must work. Nested YAML does not grep # as one token, and grep is the whole retrieval mechanism under Rule 9. # Every value is a path. The validator resolves all of them; an edge that does # not resolve is an error. That is how the graph stops rotting. edges: supersedes: This replaces that. superseded_by: The inverse. Pair them. depends_on: This breaks if that changes. serves_goal: This contributes to that outcome. references: Read-only pointer to a static document. derived_from: This content came from that source. about: Subject pointer to the canonical document on a topic. related: Untyped. Deprecated, kept so 90 existing files do not break. # Edges are REPO-LOCAL. Every value is a path inside the declaring repo, and at # level 2 it must resolve. A path into another repository cannot be checked, so # it is not an edge. Cross-repo relationships live in the manifest instead, as # `plane` / `parent_operation` / `operations` (spec 4.1 and 6.3). Those are # pointers, not paths: a consumer MUST NOT try to resolve them and MUST NOT fail # a repo it cannot reach the other end of. manifest_cross_repo: plane: "self | business. Which kind of operation this repo is." parent_operation: "On a business plane, the self plane it was deployed from." operations: "On a self plane, the business planes deployed from it." # ── Transcript and session fields ──────────────────────────────────────────── # `speakers` maps a raw body label to a real person WITHOUT touching the body. # That is deliberate: a 2026-08-04 incident was a bulk rewrite of # a customer's files. The fix that adds a mapping beats the fix that rewrites # the text. speaker_attribution: labelled: Names are already in the body (Zoom VTT). Derivable automatically. mapped: Body has anonymous labels; `speakers` carries the mapping. anonymous: Nobody knows who is who. An agent MUST refuse to attribute. transcript_required: [engagement, session_date, speaker_attribution] # ── topics: the subject layer ──────────────────────────────────────────────── # Closed, unlike `tags`. Tags record where a file came from; topics record what # is in it. Every hq-upload transcript currently carries the identical four # provenance tags, which is why nothing is findable by subject. # Seeded from tags used more than twice. Extend via .oac/manifest.yml. topics: - positioning - pricing - partnership - equity - go-to-market - marketing - content - social - sales - onboarding - delivery - curriculum - cohort - product - platform - infrastructure - security - automation - agents - claude-code - github - repo-structure - frontmatter - context-core - self-as-code - brand - website - events - aws - hiring - finance - legal - operations - tooling - data # ── Conformance: graded, so a messy repo is still usable ───────────────────── # A repo is not conformant or non-conformant. It sits at a level, and the level # says what a consumer can offer. That makes conformance a ladder a customer # climbs, not a gate that locks them out on day one with 200 legacy files. # The paid engine sits at level 3, so the ladder matches the commercial model # (ADR-090: the format is open, the engine is paid). conformance_levels: 0: name: readable requires: "Markdown in git. Frontmatter, where present, parses as YAML." unlocks: "Open, render, full-text search." 1: name: typed requires: "Every document carries title, type and audience, with legal values." unlocks: "Filter by kind, route by audience, honour the privacy gate." 2: name: linked requires: "Declared edges resolve to real paths. Directories over 10 documents carry index.md." unlocks: "Relation graph, impact analysis, progressive disclosure." 3: name: live requires: "Execution and presence conventions (see profiles.live)." unlocks: "Board, runs, collaboration. This is the paid engine surface." # ── Robustness: the rule that stops every revision breaking every customer ─── # Borrowed from OKF SPEC.md §11, and non-negotiable. A published standard whose # consumers reject unfamiliar input cannot be revised without a flag day. robustness: consumers_must_not_reject_on: - unknown type values - unknown additional frontmatter keys - broken cross-links - missing index.md - an unrecognised oac_version (attempt best-effort, never refuse) rationale: > A consumer degrades to the highest level it can satisfy and says so. It never refuses the repo. # ── Profiles: optional conventions, each unlocking specific behaviour ──────── # These are documented because they are ALREADY load-bearing in the shipped # client and were previously written down nowhere. Undocumented grammar that a # product depends on is the definition of a standard that cannot be kept. profiles: self-core: file: self-as-code.md description: "The operator's strategic identity. The one file the client parses today." frontmatter: [last_raised, next_raise] body_grammar: beats: > Level-two headings mapping to beat keys, case-insensitive, optional leading 'the': baseline, readme (also 'read me'), main, rewrite, floor, cut, commits (also 'commit'), refactor (also 'review', the heading new documents write from 1.10). invariants: "A heading 'Invariants' or 'the Invariants'." coasting: "An invariant line ending in [coasting]. The one machine-readable health signal." split: > A heading 'the split' holding one GFM table: Slice | Beat | Routing | Destination, found by header name, case-insensitively. Routing is CLOSED and bare: stays-personal, routes-to, promotable-unrouted. Destination is required when Routing is routes-to, as or /. A slice is a statement, never a file. Absent section means nothing has been routed, and everything stays personal (ADR-118 4, ADR-122 1). columns: routing: [stays-personal, routes-to, promotable-unrouted] source: products/oac-client/src/lib/parseSelfAsCode.ts reader: code/scripts/self-as-code-split.py live: description: "Execution conventions. Level 3, the paid engine surface." frontmatter: [sprint] body_grammar: items: > GFM task-list items, '- [ ]' / '- [x]' / '- [X]', also '*' bullets, at any indent. Lines inside fenced code blocks are never items. grouping: "headingPath is the nearest ancestor heading chain." deviation: > An indented blockquote directly under an item whose first line matches '> deviation (, ):'. source: products/oac-client/src/lib/live/checklist.ts pack: description: "The Context Core deep pack (ADR-094). A profile of the format, not a rival to it." files: [CLAUDE.md, README.md, CHANGELOG.md] dirs: [context-core/, decisions/, .claude/commands/] frontmatter: [version, last_updated, status] status_values: [seeded, draft, confirmed] note: > ADR-094 previously defined its own frontmatter schema. It now defers here. seeded/draft/confirmed survives as a legal pack subset because /install reads it and 172 files carry it. self-plane: description: "One person's own operation, private, above the businesses they run. ADR-118." manifest: [plane, operations] files: [self-as-code.md] dirs: [standards/, projects/, archive/, decisions/, evidence/] legacy_dirs: {history/: archive/, library/: evidence/} fields: venture: "The venture's slug. The join key: the baseline's ### block, standards/ventures//, the split's Destination and the business repo in operations. ADR-126, the Self as Code standard 4." goal: "A goal that crosses to a business repo: /. Both repos carry the same id, because an edge cannot cross repos." contact_card: file: "README.md of a folder for one person, in standards/ (from 1.13, ADR-133 Rule 4)" fields: [name, emails, phone, company, kind, crm_person, next_step] emails: "flow-style list" phone: "E.164, quoted: \"+61400000000\"" kind: [partner, customer, working-contact, friend, family] note: "Every field optional. A tool never overwrites a field a person filled in. Last contact is worked out from the folder, never a field." goal_cascade: "Goals link upward with the serves_goal edge (a path inside this repo), never a parent: key." body_grammar: spine: "self-as-code.md carries the self-core profile above." buckets: > SPADE (ADR-126, from 1.11). Filed by what you can act on, not by subject. standards/ is ongoing with a bar you hold, projects/ has an end date, archive/ is done or dropped, decisions/ holds one file per decision made, evidence/ is what you look things up in and cite. Readers accept the legacy names history/ (archive/) and library/ (evidence/); a tool never moves a person's folders for them. note: > A private repo, so every document in it is `owner` by default. This is the destination 3.5 has been sending owner documents to since 1.7. Separate repo rather than a folder because a git host grants read at the repo, so one collaborator on one business would otherwise see the operator's identity and every other business. Lineage: filing by actionability is PARA, by Tiago Forte; SPADE is the same idea with a name you can say. systems: description: "What an operation runs on today. The counterpart to a Context Core. ADR-118." manifest: [plane] files: [systems/register.md] body_grammar: sections: > Up to three level-two sections, each holding one GFM table, all optional: '## Software' (one row per tool), '## Where the work lives' (one row per place files or records sit), '## Processes' (one row per repeated thing). A missing section means it was not captured, never an error. closed_columns: > Found by header name, case-insensitively. A consumer ignores columns it does not recognise; every other column is free text. A closed cell carries the BARE value: no bold, no code ticks, no trailing note. '| manual |', never '| **manual** |'. Emphasis defeats the grep the whole profile rests on. columns: connects_via: [api, mcp, scrape, none] runs: [manual, semi, automated] call: [keep, automate, build, replace, retire] note: > Rows because the point is counting: `Runs` turns 'we should automate more' into a number that moves, and both closed columns answer to `grep -c`, which is the only tooling this profile assumes. The reader is code/scripts/systems-register.py in the reference implementation; it checks the closed cells and counts the baseline. No logos, no per-app files, no frontmatter schema (ADR-107, and the scalar-reader reason behind array_style: flow).