Skip to content

claw: TRI CLAW social-visual skill cards -- format, scene, uniqueness gate, publish consent (Closes #10570) - #10574

Merged
gHashTag merged 2 commits into
claude/t27f-content-factoryfrom
claude/social-visual-cards-10570
Oct 18, 2026
Merged

gHashTag merged 2 commits into
claude/t27f-content-factoryfrom
claude/social-visual-cards-10570

Conversation

@gHashTag

Copy link
Copy Markdown
Owner

Closes #10570. Sub-issue R5 of epic #10560. Stacked on #10566 (reel_factory.t27, not yet on master): the base is that PR's branch, so this diff is only R5. Retarget to master once #10566 merges.

TRI CLAW now has t27 skill cards for making social-media visuals from the content factory's plan, so it stops choosing sizes and scenes by judgement and cannot publish a near-duplicate or an unapproved post.

What is added

  • specs/claw/social_visuals.t27 -- the four decisions, once, with 14 tests. It imports reel_factory (formats, arche, too_similar, carousel_slides) and reel_topics::TOPICS; no table is copied.
    • format: pick_format(channel, kind) over 8 channels and 5 kinds; a moving format only for a moving kind; FMTS ("unsupported") for a pair with no format.
    • scene: scene_for(t) is reel_factory::arche; an item outside the course gets ARCHS (no plan).
    • gate: gate_verdict(items, ran, pairs, worst); blocks an empty, unmeasured, incomplete or too-alike set.
    • consent: publish_verdict(...); allowed only with a passed gate and the owner's yes, from the owner's chat, for the same post digest and channel; a digest of 0 approves nothing.
  • specs/skills/t27-social-visual-{format,scene,gate,consent}.t27 -- schema-only cards per specs/skills/README.md, each with a basic_constants test, each naming the law and reel_factory in SPECS.
  • 5 seals and 5 t27b ledger rows (max_not_pass unchanged).

Lab results (Railway lab, nothing run on the Mac)

  • t27c test-report: law 14/14 pass, 0 vacuous; each card 1/1 pass, 9 asserts, 0 vacuous.
  • t27c seal --save then --verify: all five "all hashes MATCH".
  • t27b build --elf: all five (and reel_factory) produce an object, rc 0.
  • Hand-written code outside t27: 0 lines.

Please look at

  • t27b JIT verdict is not run here. The lab is x86 and t27b test refuses it (the JIT is arm64 only), the same as for t27f: reel_factory.t27 -- content features to a unique scene plan per item (Closes #10561) #10566. The five ledger rows say pass on the strength of the front end and the object build; the arm64 ratchet is the check. If it disagrees, change the row.
  • Cards have SOURCE = "". There is no SKILL.md behind them, so the site will label them spec-only (a recognised witness label). Say so if you want .claude/skills/ bodies written as a follow-up.
  • Chosen, not verified: the channel-to-format table is this module's choice among the factory's eight formats, not read from each network's current rules; the 850 permille bound is still the factory's unmeasured starting value; the gate compares stills within one run only; only the 27 course lessons have plans; digest computation and "did this come from the owner's chat" are the publisher host's job. Publishing itself is out of scope. These are stated in the spec header.

🤖 Generated with Claude Code

gHashTag and others added 2 commits October 19, 2026 00:40
#10570)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Closes #10570)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@gHashTag
gHashTag merged commit 10cd62d into claude/t27f-content-factory Oct 18, 2026
32 of 35 checks passed
gHashTag added a commit that referenced this pull request Oct 18, 2026
… gate, publish consent (Closes #10570) (#10574) (#10613)

* claw: social_visuals.t27 and four t27-social-visual skill cards (Closes #10570)



* seals and t27b ledger rows for social_visuals and its four skill cards (Closes #10570)



---------


(cherry picked from commit 10cd62d)

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant