@profile-as-code/cli 0.3.6 → 0.4.0

package/skills/medium/SKILL.md

@@ -0,0 +1,86 @@
+---
+name: medium-post
+description: >
+  Long-form post skill for Medium. Says what a Medium essay is for and how the
+  agent writes it; the machine contract is
+  skills/medium/longform.yaml (required/optional fields, editorial word band,
+  export: api). The judgment layer — shape, voice, GEO, i18n — lives in post.md.
+  The agent writes N variants, the operator picks one, and the choice is recorded
+  with `pasco post choose`. The contract's api export is mapped, not guaranteed:
+  verify API access before relying on it.
+model: juicio (frontier) — the position and its stakes are judgment; the
+  mechanical parts (markdown, front matter) are not
+applies_to: medium
+projection: agent-agnostic (any harness that reads Markdown and can run the CLI)
+version: 0.1.0
+status: draft (#361)
+sources: skills/medium/longform.yaml
+---
+
+# Medium long-form — skill
+
+## Mission
+
+Turn one canonical seed into an essay that argues a position a professional
+audience can disagree with: the situation, the tension, what was done, what it
+cost. The skill is the judgment layer (D5); the contract plus `pasco post check`
+are the deterministic part.
+
+## When it runs
+
+`pasco post task <seed> --format longform --channel medium` emits a task bundle
+that points at the contract (`longform.yaml`) and the judgment file (`post.md`);
+it does **not** carry this `SKILL.md`. This file is the channel's strategy
+reference for whoever opens it directly: read the contract for the fields and the
+band, then `post.md` for how to write.
+
+Do **not** run `pasco refine medium` for this channel. Refine is the profile
+command: with a `SKILL.md` now present it resolves this file instead of answering
+"no skill for medium yet", and emits a **profile** task ("write N variants of the
+medium profile") that saves to `profiles/medium/variants/<id>.md` — not a post
+workflow. Telling a profile skill from a post skill is an engine decision, not
+something this skill can fix.
+
+## Flow
+
+1. Read `longform.yaml` + `post.md`; the contract owns every limit.
+2. Write N variants as YAML front matter + markdown body, at
+   `posts/<seed>/medium/medium-<n>.md`.
+3. The operator picks one; record it with `pasco post choose`.
+4. `pasco post check` — fix the findings.
+5. `pasco post export <seed> --target postiz` for the API draft payload.
+6. Publish manually and verify the live page (title, tags, canonical).
+
+## Contract pointers (single source each)
+
+| Question | Home |
+|---|---|
+| required / optional fields | `longform.yaml` `fields` |
+| editorial word band | `longform.yaml` `limits.words` (not a platform cap) |
+| api vs paste | `longform.yaml` `export` |
+| link policy | `longform.yaml` `links.policy` (`identity-ssot-only`) |
+| dated platform sources | `longform.yaml` `sources[]` |
+| how to write the piece | `post.md` |
+
+## Open precondition
+
+The contract maps `export: api` onto the Medium posting API, but whether an
+operator can still obtain write access is not stated by the contract's source:
+treat it as unverified [?] and check before promising an API publish.
+
+## GEO and i18n
+
+Public, indexed essay pages; tags are browsing paths, not search intent; the
+citation unit is the stated position plus the evidence under it. Write `en` by
+default; a Spanish essay is a separate seed for that audience, never a translation
+sharing a canonical. The crawler stance is the repo's global `ai.policy`
+(`platforms/registry.yaml`, v1, 2026-10-27).
+
+## Verification
+
+- `pasco post check` exit 0 (no `missing-field`, `word-count` or
+  `url-not-in-ssot` errors).
+- The title is arguable; the dek states the stake.
+- Facts byte-identical to the SSOT; canonical (when set) points at the SSOT home.
+- Report any contract gap as a finding — the engine is not extended from a
+  skill (#361).

package/skills/medium/longform.yaml

@@ -0,0 +1,33 @@
+# medium long-form rules — machine-readable (#266)
+#
+# Read by `pasco post task|check|export` when the seed is `format: longform`.
+# The word range is an EDITORIAL default (the operator's house style), NOT a
+# platform limit: cite a platform source before encoding a hard cap here.
+
+platform: medium
+format: longform
+export: api
+settings:
+  __type: medium
+fields:
+  required:
+    - title
+    - dek
+  optional:
+    - tags
+    - canonical
+    - publication
+limits:
+  words:
+    min: 600
+    max: 3000
+links:
+  policy: identity-ssot-only
+structure:
+  - title
+  - dek
+  - sections
+sources:
+  - url: https://github.com/Medium/medium-api-docs
+    date: 2026-10-29
+    note: "Medium's posting API takes the title, tags and canonical URL in the post payload; the body is markdown."

package/skills/medium/post.md

@@ -0,0 +1,96 @@
+# Medium long-form post — skill (MVP)
+
+> Long-form post skill. The contract is `longform.yaml`: required/optional
+> fields, the editorial word band and `export` live there and are **not**
+> restated here — this file is the judgment layer the agent reads before
+> writing. The contract's `sources[]` carries the dated URL for every platform
+> claim; re-verify before making a hard one, and cite a source (date + tier)
+> before calling anything a limit.
+
+## What a Medium piece is for
+
+An essay for a professional audience that reads to *understand a position*, not
+to copy a snippet. Medium is a reading surface with tags and publications as its
+two distribution paths; the piece competes for a reader's attention against other
+essays, so the argument — not the feature list — is what travels. The technical
+depth lives on the other channels; here the decision, its cost and its
+consequence carry the piece.
+
+## Shape
+
+1. **Title** — the position in one line. Concrete and arguable; if nobody could
+   disagree, it is not a title for this channel.
+2. **Dek** — the stake: what the reader gets by finishing. The contract requires
+   it as a field and the engine delivers it as front matter, never as the body's
+   first lines. No platform mapping is declared; confirm on the first publish how
+   Medium renders it [?].
+3. **Sections** — narrative arc, not a manual: situation → tension → what was
+   done → what it cost → what follows. H2 per move; short paragraphs; one idea
+   per paragraph.
+
+## Rules
+
+- **Front matter is the field contract.** `longform.yaml` declares the fields this
+  channel consumes, required and optional alike; the engine exports only those,
+  and an undeclared field is dropped (`pasco post check` warns `unknown-field`) —
+  never smuggle one into the body instead.
+- **Length:** the band in `limits.words` is the operator's editorial default (the
+  contract says so), not a Medium limit — Medium does not publish a cap this
+  contract relies on.
+- **Links:** `links.policy` is `identity-ssot-only`. When the piece is syndicated,
+  `canonical` points at the canonical home from the SSOT — a Medium URL is never
+  the canonical for the original.
+- **Facts:** `{{fact:<pointer>}}` for every number, date, URL and claim of
+  achievement.
+- **Voice:** first person, essayistic, no hype and no "lessons learned" filler.
+  Match `brandVoice`; Medium's audience rewards specificity, not enthusiasm.
+
+## Export: `api`
+
+The contract declares `export: api`: the Medium posting API takes the title, tags
+and canonical URL in the post payload and the body as markdown [OFICIAL — the
+contract's source, `sources[]` in `longform.yaml`, dated 2026-10-29]. Two
+consequences:
+
+- The writer must not depend on a formatting feature the API payload does not
+  carry: markdown, no raw HTML.
+- Whether the operator can still obtain write access to that API is **not**
+  stated by the contract's source: verify access before relying on `export: api`
+  for a real publish [?]. The export itself only produces a Postiz draft payload,
+  and publishing stays manual (no scheduler, no OAuth — `docs/post-mvp.md`).
+
+## GEO (how this channel is read and cited)
+
+Medium pages are public and indexed, and the tags are browsing paths rather than
+search intent. The quotable unit is a stated position plus the evidence
+immediately under it, so keep the claim next to its proof instead of building a
+reveal. The crawler stance is the repo's global `ai.policy`
+(`platforms/registry.yaml`, v1, 2026-10-27); no per-channel rule is invented
+here. Consistency check: every fact agrees with the profile and with the other
+channels' projections, byte for byte.
+
+## i18n
+
+Write `en` by default: the audience and the tag vocabulary are English-first, and
+a publication is per language. A Spanish essay is its own seed for that audience,
+not a translated copy of the English one — two canonicals of one text split the
+citation.
+
+## Anti-patterns
+
+- A "10 lessons" listicle with no evidence per lesson.
+- An essay that opens with a dictionary definition.
+- A `canonical` invented for the occasion.
+- Syndicating the original without a canonical back to the SSOT home.
+- Enthusiasm in place of an argument.
+
+## Output contract
+
+Save each variant as YAML front matter + a markdown body at
+`posts/<seed>/medium/medium-<n>.md`, then record the choice and check it:
+
+```
+pasco post choose <seed> --channel medium --variant medium-1
+pasco post check <seed>     # missing-field · word-count · url-not-in-ssot · unknown-field
+pasco post export <seed> --target postiz
+```

package/skills/substack/SKILL.md

@@ -0,0 +1,83 @@
+---
+name: substack-post
+description: >
+  Long-form post skill for Substack. Says what a Substack issue is for and how
+  the agent writes it; the machine contract is
+  skills/substack/longform.yaml (required/optional fields, editorial word band,
+  export: paste). The judgment layer — shape, voice, GEO, i18n — lives in
+  post.md. The agent writes N variants, the operator picks one, and the choice is
+  recorded with `pasco post choose`; the deploy is a manual paste driven by the
+  generated checklist, because Substack exposes no body-publish API.
+model: juicio (frontier) — the letter's opening and its closing are judgment; the
+  mechanical parts (markdown, front matter, paste checklist) are not
+applies_to: substack
+projection: agent-agnostic (any harness that reads Markdown and can run the CLI)
+version: 0.1.0
+status: draft (#361)
+sources: skills/substack/longform.yaml
+---
+
+# Substack long-form — skill
+
+## Mission
+
+Turn one canonical seed into a newsletter issue that works twice: as the email a
+subscriber opens and as the public archive page a visitor lands on. One idea, a
+stake, an argument developed once, and a closing that says what comes next. The
+skill is the judgment layer (D5); the contract plus `pasco post check` are the
+deterministic part.
+
+## When it runs
+
+`pasco post task <seed> --format longform --channel substack` emits a task bundle
+that points at the contract (`longform.yaml`) and the judgment file (`post.md`);
+it does **not** carry this `SKILL.md`. This file is the channel's strategy
+reference for whoever opens it directly: read the contract for the fields and the
+band, then `post.md` for how to write.
+
+Do **not** run `pasco refine substack` for this channel. Refine is the profile
+command: with a `SKILL.md` now present it resolves this file instead of answering
+"no skill for substack yet", and emits a **profile** task ("write N variants of
+the substack profile") that saves to `profiles/substack/variants/<id>.md` — not a
+post workflow. Telling a profile skill from a post skill is an engine decision,
+not something this skill can fix.
+
+## Flow
+
+1. Read `longform.yaml` + `post.md`; the contract owns every limit.
+2. Write N variants as YAML front matter + markdown body, at
+   `posts/<seed>/substack/substack-<n>.md`.
+3. The operator picks one; record it with `pasco post choose`.
+4. `pasco post check` — fix the findings.
+5. `pasco post export <seed>` → `paste-checklist.md` (no body API in v1).
+6. Paste the body and each field into the editor, tick the checklist, and verify
+   the live archive page before calling it published.
+
+## Contract pointers (single source each)
+
+| Question | Home |
+|---|---|
+| required / optional fields | `longform.yaml` `fields` |
+| editorial word band | `longform.yaml` `limits.words` (not a platform cap) |
+| api vs paste | `longform.yaml` `export: paste` (no body API) |
+| link policy | `longform.yaml` `links.policy` (`identity-ssot-only`) |
+| dated platform sources | `longform.yaml` `sources[]` |
+| how to write the issue | `post.md` |
+
+## GEO and i18n
+
+The archive page is public and indexable while the email copy is not; the section
+heading plus its claim is the citation unit, and it resolves to the archive URL.
+Keep facts byte-identical to the profile. One publication per language; `en` by
+default, a Spanish publication as its own surface. The crawler stance is the
+repo's global `ai.policy` (`platforms/registry.yaml`, v1, 2026-10-27).
+
+## Verification
+
+- `pasco post check` exit 0 (no `missing-field`, `word-count` or
+  `url-not-in-ssot` errors).
+- The pasted body kept headings, links and quotes; the archive page renders the
+  dek and title as intended.
+- Facts byte-identical to the SSOT; every URL from the identity SSOT.
+- Report any contract gap as a finding — the engine is not extended from a
+  skill (#361).

package/skills/substack/longform.yaml

@@ -0,0 +1,36 @@
+# substack long-form rules — machine-readable (#266)
+#
+# Read by `pasco post task|check|export` when the seed is `format: longform`.
+# The word range is an EDITORIAL default (the operator's house style), NOT a
+# platform limit: cite a platform source before encoding a hard cap here.
+# `export: paste` because Substack exposes no body-publish API — the post is
+# pasted into the editor (#266 decision).
+
+platform: substack
+format: longform
+export: paste
+settings:
+  __type: substack
+fields:
+  required:
+    - title
+    - dek
+  optional:
+    - preview
+    - tags
+    - canonical
+    - list
+limits:
+  words:
+    min: 600
+    max: 3000
+links:
+  policy: identity-ssot-only
+structure:
+  - title
+  - dek
+  - sections
+sources:
+  - url: https://support.substack.com/
+    date: 2026-10-29
+    note: "Substack publishes from its editor and has no public post-publish API; v1 treats it as a paste surface (#266)."

package/skills/substack/post.md

@@ -0,0 +1,91 @@
+# Substack long-form post — skill (MVP)
+
+> Long-form post skill. The contract is `longform.yaml`: required/optional
+> fields, the editorial word band and `export` live there and are **not**
+> restated here — this file is the judgment layer the agent reads before
+> writing. The contract's `sources[]` carries the dated URL for every platform
+> claim; re-verify before making a hard one, and cite a source (date + tier)
+> before calling anything a limit.
+
+## What a Substack piece is for
+
+A newsletter issue for subscribers that also becomes a public archive page. Two
+readers, one text: the subscriber who opens it in an inbox and the visitor who
+lands on the archive from a link. Substack is a reading surface with a
+subscription relation attached — the issue is a letter that ends with a reason to
+keep receiving it, not a post that closes with a hashtag.
+
+## Shape
+
+1. **Title** — the line that earns the open in an inbox and the click in search.
+   State the value of reading, not just the topic.
+2. **Dek** — the line under the title in the archive and the setup in the email:
+   one sentence of stake. The contract requires it; how the publication renders
+   it is editor-controlled [?] — verify on the first publish.
+3. **Sections** — the letter opens with the point, develops it once, and closes
+   with what comes next. H2 per move; the closing line is written, never left to
+   "thanks for reading".
+
+## Rules
+
+- **Front matter is the field contract.** `longform.yaml` declares the fields this
+  channel consumes, required and optional alike; the engine exports only those,
+  and an undeclared field is dropped (`pasco post check` warns `unknown-field`) —
+  never smuggle one into the body instead.
+- **Length:** the band in `limits.words` is the operator's editorial default (the
+  contract says so), not a Substack limit. It is the budget for an issue read in
+  one sitting.
+- **Links:** `links.policy` is `identity-ssot-only`. `canonical` is only set when
+  the SSOT already carries the target URL; the archive page is a public URL of
+  its own and, once live, is what other channels can point at — but that pointer
+  enters the SSOT first, never the prose.
+- **Facts:** `{{fact:<pointer>}}` for every number, date and URL.
+- **Voice:** first person, letter-like, no hype. Match `brandVoice`; a newsletter
+  is where the operator's actual voice is the differentiator.
+
+## Export: `paste`
+
+The contract declares `export: paste` because Substack publishes from its editor
+and exposes no body-publish API [OFICIAL — the contract's source, `sources[]` in
+`longform.yaml`, dated 2026-10-29]. Consequences for the writing and the deploy:
+
+- The body is **pasted into the editor**; keep the markdown paste-safe
+  (headings, paragraphs, lists, bold/italic, block quotes, links). Re-check
+  formatting after pasting — nothing should arrive flattened.
+- Fields are pasted one by one; the generated `paste-checklist.md` is the record
+  of what was pasted and in what order.
+- The export is a checklist, not a publish: sending stays a manual action.
+
+## GEO (how this channel is read and cited)
+
+The archive page is public and indexable; the email copy is not. The citation
+unit is a section heading plus the claim under it, resolving to the archive URL —
+which is why the piece must read as well on the web as in the inbox. The crawler
+stance is the repo's global `ai.policy` (`platforms/registry.yaml`, v1,
+2026-10-27); no per-channel rule is invented here. Consistency check: the issue's
+facts are byte-identical to the profile and to the other channels' projections.
+
+## i18n
+
+A publication is per language: one publication, one subscriber base, one
+language. Write `en` by default; a Spanish publication is a separate surface with
+its own archive. Do not send the same text in two languages to one list.
+
+## Anti-patterns
+
+- An issue that opens with "welcome back" instead of the point.
+- A dek that repeats the title with different words.
+- A "read the full piece on…" issue that is only a teaser.
+- Rich formatting that the editor flattens on paste.
+- A closing that asks for shares but not for a reply or a next step.
+
+## Output contract
+
+Save each variant as YAML front matter + a markdown body at
+`posts/<seed>/substack/substack-<n>.md`, then record the choice and check it:
+
+```
+pasco post choose <seed> --channel substack --variant substack-1
+pasco post check <seed>     # missing-field · word-count · url-not-in-ssot · unknown-field
+pasco post export <seed>    # paste: writes posts/<seed>/substack/paste-checklist.md
+```

package/skills/wordpress/SKILL.md

@@ -0,0 +1,84 @@
+---
+name: wordpress-post
+description: >
+  Long-form post skill for WordPress. Says what a WordPress post is for and how
+  the agent writes it; the machine contract is
+  skills/wordpress/longform.yaml (required/optional fields, editorial word band,
+  export: api). The dek→excerpt mapping is NOT declared by the contract: it is an
+  open question [?], not a promise. The judgment layer — shape,
+  voice, GEO, i18n — lives in post.md. The agent writes N variants, the operator
+  picks one, and the choice is recorded with `pasco post choose`.
+model: juicio (frontier) — what the page answers and how it is structured is
+  judgment; the mechanical parts (markdown, front matter) are not
+applies_to: wordpress
+projection: agent-agnostic (any harness that reads Markdown and can run the CLI)
+version: 0.1.0
+status: draft (#361)
+sources: skills/wordpress/longform.yaml
+---
+
+# WordPress long-form — skill
+
+## Mission
+
+Turn one canonical seed into a durable page on a site the operator owns: the
+question answered first, the reasoning under it, the caveats at the end, and a
+heading structure that a reader or an extractor can navigate. The skill is the
+judgment layer (D5); the contract plus `pasco post check` are the deterministic
+part.
+
+## When it runs
+
+`pasco post task <seed> --format longform --channel wordpress` emits a task bundle
+that points at the contract (`longform.yaml`) and the judgment file (`post.md`);
+it does **not** carry this `SKILL.md`. This file is the channel's strategy
+reference for whoever opens it directly: read the contract for the fields and the
+band, then `post.md` for how to write.
+
+Do **not** run `pasco refine wordpress` for this channel. Refine is the profile
+command: with a `SKILL.md` now present it resolves this file instead of answering
+"no skill for wordpress yet", and emits a **profile** task ("write N variants of
+the wordpress profile") that saves to `profiles/wordpress/variants/<id>.md` — not
+a post workflow. Telling a profile skill from a post skill is an engine decision,
+not something this skill can fix.
+
+## Flow
+
+1. Read `longform.yaml` + `post.md`; the contract owns every limit.
+2. Write N variants as YAML front matter + markdown body, at
+   `posts/<seed>/wordpress/wordpress-<n>.md`.
+3. The operator picks one; record it with `pasco post choose`.
+4. `pasco post check` — fix the findings.
+5. `pasco post export <seed> --target postiz` for the API draft payload.
+6. Publish manually and verify the live page (title, dek, tags, canonical).
+
+## Contract pointers (single source each)
+
+| Question | Home |
+|---|---|
+| required / optional fields | `longform.yaml` `fields` |
+| editorial word band | `longform.yaml` `limits.words` (not a platform cap) |
+| api vs paste | `longform.yaml` `export` |
+| link policy | `longform.yaml` `links.policy` (`identity-ssot-only`) |
+| dated platform sources | `longform.yaml` `sources[]` |
+| how to write the page | `post.md` |
+
+## GEO and i18n
+
+A page on the operator's own domain: public, indexed, stable — the most citable
+surface of the set. The headings are the outline, the claim sits directly under
+the heading that names it, and the dek is a field whose rendering the contract
+leaves unmapped [?]. Write `en` by default; a
+Spanish post belongs to the Spanish site with its own URL. The crawler stance is
+the repo's global `ai.policy` (`platforms/registry.yaml`, v1, 2026-10-27).
+
+## Verification
+
+- `pasco post check` exit 0 (no `missing-field`, `word-count` or
+  `url-not-in-ssot` errors).
+- Heading hierarchy is structural (H2 → H3), not visual.
+- The dek→excerpt mapping is unverified [?]: confirm on the first publish whether
+  the theme's excerpt is the dek or a separate field.
+- Facts byte-identical to the SSOT; every URL from the identity SSOT.
+- Report any contract gap as a finding — the engine is not extended from a
+  skill (#361).

package/skills/wordpress/longform.yaml

@@ -0,0 +1,33 @@
+# wordpress long-form rules — machine-readable (#266)
+#
+# Read by `pasco post task|check|export` when the seed is `format: longform`.
+# The word range is an EDITORIAL default (the operator's house style), NOT a
+# platform limit: cite a platform source before encoding a hard cap here.
+
+platform: wordpress
+format: longform
+export: api
+settings:
+  __type: wordpress
+fields:
+  required:
+    - title
+    - dek
+  optional:
+    - tags
+    - canonical
+    - publication
+limits:
+  words:
+    min: 600
+    max: 3000
+links:
+  policy: identity-ssot-only
+structure:
+  - title
+  - dek
+  - sections
+sources:
+  - url: https://developer.wordpress.org/rest-api/reference/posts/
+    date: 2026-10-29
+    note: "The WordPress REST posts endpoint takes title, content, tags and an excerpt."

package/skills/wordpress/post.md

@@ -0,0 +1,99 @@
+# WordPress long-form post — skill (MVP)
+
+> Long-form post skill. The contract is `longform.yaml`: required/optional
+> fields, the editorial word band and `export` live there and are **not**
+> restated here — this file is the judgment layer the agent reads before
+> writing. The contract's `sources[]` carries the dated URL for every platform
+> claim; re-verify before making a hard one, and cite a source (date + tier)
+> before calling anything a limit.
+
+## What a WordPress piece is for
+
+The piece on a site the operator controls: a durable page with its own URL,
+its own theme and its own permanence. WordPress supports the longest shelf life
+of any channel here — a well-structured post keeps being found and cited after
+the feeds move on — so write for the reader who arrives six months later with a
+question, not only for the reader who subscribes today.
+
+## Shape
+
+1. **Title** — the question the post answers or the outcome it delivers, in the
+   words a reader would search.
+2. **Dek** — the summary. The contract requires it as a field and declares no
+   mapping to the REST `excerpt`: the endpoint takes an excerpt per the
+   contract's source, but whether the engine's `dek` arrives there is unverified
+   [?]. Write it as a standalone summary and confirm the mapping on the first
+   publish.
+3. **Sections** — one H2 per step or idea, H3 only for sub-steps; the heading
+   hierarchy is the page's table of contents and its extractable structure. Open
+   with the answer, then the reasoning, then the caveats.
+
+## Rules
+
+- **Front matter is the field contract.** `longform.yaml` declares the fields this
+  channel consumes, required and optional alike; the engine exports only those,
+  and an undeclared field is dropped (`pasco post check` warns `unknown-field`) —
+  never smuggle one into the body instead.
+- **Length:** the band in `limits.words` is the operator's editorial default (the
+  contract says so), not a WordPress limit. It is the budget for a page that
+  answers one question completely.
+- **Links:** `links.policy` is `identity-ssot-only`. On a site the operator owns,
+  internal links are still SSOT links — do not type a URL the SSOT does not
+  carry, and set `canonical` only when the piece is syndicated and the SSOT
+  carries the original.
+- **Facts:** `{{fact:<pointer>}}` for every number, date and URL.
+- **Voice:** first person, explanatory, no hype. Match `brandVoice`; a durable
+  page is written to be re-read, so avoid references that only make sense "this
+  week" unless the date is stated.
+
+## Export: `api`
+
+The contract declares `export: api`: the WordPress REST posts endpoint takes the
+title, content, tags and an excerpt [OFICIAL — the contract's source, `sources[]`
+in `longform.yaml`, dated 2026-10-29]. Consequences for the writing:
+
+- The body is content the endpoint receives; keep it valid markdown/HTML with a
+  clean heading hierarchy — no layout that depends on the editor's blocks.
+- The contract declares no dek→excerpt mapping. The endpoint takes an excerpt per
+  its source, and whether the engine's `dek` lands there is unverified [?] —
+  confirm on the first publish. Either way the dek is a field, never the body's
+  first paragraph.
+- The export produces a Postiz draft payload; publishing stays manual (no
+  scheduler, no OAuth — `docs/post-mvp.md`).
+
+## GEO (how this channel is read and cited)
+
+A page on the operator's own domain is the most citable surface here: public,
+indexed, stable URL. Structure is what gets extracted — a question-shaped title, a
+summary in the dek field (mapped to the excerpt only if the site does that, [?]),
+and claims stated directly under the heading that
+names them. The crawler stance is the repo's global `ai.policy`
+(`platforms/registry.yaml`, v1, 2026-10-27); no per-channel rule is invented
+here. Consistency check: every fact matches the profile byte for byte, and the
+page's dates stay truthful as it ages.
+
+## i18n
+
+A WordPress site is per language (one site, one locale, or a multilingual setup
+the operator already runs). Write `en` by default; a Spanish post belongs to the
+Spanish site, and each translation is its own post with its own URL — never a
+duplicate of the English one under the same canonical.
+
+## Anti-patterns
+
+- A page that buries the answer under three paragraphs of context.
+- Heading levels used for visual size instead of structure.
+- A dek repeated verbatim as the first paragraph.
+- Time-bound language with no date ("recently", "this week").
+- Syndicated content with no canonical back to the original.
+
+## Output contract
+
+Save each variant as YAML front matter + a markdown body at
+`posts/<seed>/wordpress/wordpress-<n>.md`, then record the choice and check it:
+
+```
+pasco post choose <seed> --channel wordpress --variant wordpress-1
+pasco post check <seed>     # missing-field · word-count · url-not-in-ssot · unknown-field
+pasco post export <seed> --target postiz
+```