Chatgpt Finalize

• one markdown handoff artifact containing PRIMARY_ARTIFACT, PREFILLED_PACKAGING_TEMPLATE, NEXT_STEP_DESTINATION, NEXT_STEP_REQUIRED_INPUTS, Validation Notes, and NEXT_ACTION
• include TOOL_PROMPT_BLOCK only when needed to carry package-safe article payload forward without asking post-package to invent content
• if some packaging inputs remain unsafe to infer, leave only those fields blank inside PREFILLED_PACKAGING_TEMPLATE

Active Anti-Commodity Enforcement

• preserve article substance
• preserve source lock
• flag or lightly tighten obvious AI-slop patterns only when this can be done without changing meaning
• never genericize the article
• never polish away tension, contradiction, or uncertainty
• preserve specific nouns and source-strength cautions
• if fixing AI-slop would require changing substance, stop and report rather than rewrite

Procedure

1. Confirm source lock explicitly.
2. Confirm the Claude article is the source of truth for article substance.
3. Compare the Claude article to docs/editorial/STA_POST_STANDARD.md and docs/site/POST_SHAPE.md.
4. Add only the required structure, placeholders, and Claims & Verification scaffolding needed to match the standard.
5. Preserve the Claude article's claims, meaning, examples, recommendations, and uncertainty.
6. Do not add, remove, strengthen, soften, summarize, reinterpret, or silently reorder substantive content to change meaning.
7. Apply the Active Anti-Commodity Enforcement rules directly during finalization.
8. You may tighten obvious AI-slop patterns only when this can be done without changing substance, evidence boundaries, or tone intent. Remove filler transitions, generic openings, and commodity phrasing only if the meaning remains identical.
9. Preserve markdown links exactly where they already exist. If an internal link target is unknown, keep a clear placeholder instead of guessing.
10. Keep canonical URL handling exact and package-safe only.
11. Use upstream CLAIM_SUPPORT_MAP or equivalent structured source support when it is present.
12. Stop if any required claim lacks citation support.
13. If fixing AI-slop would require new facts, stronger claims, removed tension, or softened uncertainty, stop and report instead of rewriting.
14. Build the PRIMARY_ARTIFACT as the finalized article for repo control.
15. When finalization succeeds, derive best-effort packaging proposals from available context, including article title, topic, pillar, description candidate, likely tags, impact score proposal, hero image filename proposal, hero image alt proposal, canonical URL proposal when the site pattern is known, draft and featured defaults, evergreen default, affiliate offer default, related-post placeholder, asset-reference placeholders, and destination-path proposal.
16. Build those values into PREFILLED_PACKAGING_TEMPLATE instead of a blank packaging template.
17. If a value cannot be safely inferred, leave that field blank while keeping the rest prefilled.
18. Proposed packaging values remain unapproved until the operator explicitly replies with accept or accept with edits:.
19. If useful for downstream convenience, write TOOL_PROMPT_BLOCK for post-package so it carries the finalized article markdown, Claims & Verification block, citation status, and any already-known approved packaging values, but do not require post-package to infer missing metadata from it.
20. Ask for manual citation-note reconstruction only if the upstream structured claim support is missing or incomplete.
21. If blocked because source-lock inputs or citations are incomplete or invalid, point NEXT_STEP_DESTINATION to the corrective upstream step instead of post-package.
22. The handoff to post-package must prepare package work only and must not ask packaging to write or infer article content.
23. Format the result using the standard handoff shape.
24. End with NEXT_ACTION telling the operator which next skill to run, which block or template to use, where to use it, and that they may reply with accept or accept with edits: for packaging proposals.

Acceptance Checks

• article substance matches Claude's draft
• docs/editorial/STA_POST_STANDARD.md is enforced
• active anti-commodity enforcement is applied without breaking source lock
• required sections are present
• obvious commodity filler is removed only when substance is unchanged
• Claims & Verification reflects existing claims only
• media placeholders use the documented pattern
• source lock is stated explicitly
• no uncited claims were packaged
• upstream structured claim support is used when available
• PREFILLED_PACKAGING_TEMPLATE reduces packaging-input friction by proposing safe values from available context
• proposed packaging values remain clearly unapproved until operator acceptance
• TOOL_PROMPT_BLOCK, when present, prepares package work only and does not ask packaging to infer missing metadata
• markdown links are preserved or clearly marked as placeholders
• next action is explicit and mechanical

Failure Conditions

• requested changes require substantive rewrite
• source lock is not active or the source of truth is unclear
• citations are missing
• upstream structured claim support existed but was ignored
• the draft blends the two lanes unclearly
• enforcing the standard would require new facts, new arguments, or stronger claims
• fixing AI-slop would require substantive rewriting or loss of specific source-strength cautions
• the output introduces new facts, examples, recommendations, summaries, or stronger framing
• TOOL_PROMPT_BLOCK asks packaging to invent metadata or write article content
• PREFILLED_PACKAGING_TEMPLATE silently treats proposed values as approved
• PREFILLED_PACKAGING_TEMPLATE leaves safely inferable fields blank without reason
• packaging succeeds but the handoff still requires the operator to reconstruct article payload from scattered prior sections
• link cleanup changes article substance or guesses unknown internal links
• the next action is missing or ambiguous