Claude Skill

reality-show

Keep serialized build-in-public content continuous: read one canon file of season, arcs, beats and loops and say which narrative beat to advance next, so episodes chain into a series. Pairs with the episode adapter, which does the writing. Triggers: "/reality-show", "season state

LLM Mart · 0 points · 1 views 0 listing impressions 0 install-command copies
Download tonydzi-second-brain-starter-kit-skills_reality-show-7bc2efa.zip · 3 KB

Install

skills CLI npx skills add https://github.com/tonydzi/second-brain-starter-kit/tree/master/skills/reality-show
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install tonydzi-second-brain-starter-kit@llmmart
Git git clone https://github.com/tonydzi/second-brain-starter-kit.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole tonydzi/second-brain-starter-kit collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

REALITY-SHOW v2 — the season continuity layer on top of a SINGLE CANON

GOAL: make the stream of episodes read like a SERIES. /episode writes one episode with dramaturgy (palette [[style-reality-show]]); THIS skill remembers the SEASON and suggests the next narrative move.

SOURCE OF TRUTH = the canon: $OBSIDIAN_VAULT/04-Projects\show-canon\ (season-01.md · arcs\ · beats\ · loops\ · rules; the map is _SHOW-CANON.md). ⛔ season-state.json is FROZEN as of 2026-07-10 (_imports\content-factory\season-state.json + season_state.py) — it was the third diverging spine after THREADS/SHOW-STATE; decision decision-single-canon-story-state. Do NOT read it as truth, do NOT update it.

MODES

status (default, read-only, 0 LLM tokens)

Read season-01.md (the season question, active arcs, open loops, the board) + the headers of arcs\*.md (status, current_state) + loops\*.md (open?) → a one-line digest for the owner: the season + arcs with their stakes + dangling cliffhanger loops + story_day. Nothing is changed.

next — propose the beat for the next episode (authorial model, cheap)

  1. Read: season-01.md + open loops + fresh beats in beats\ (happened, latest by occurred_on) + the 📝 inbox _imports\content-factory\triage\posts.md (fresh post material). Do NOT read the whole corpus.
  2. Judgement (top-tier model): which OPEN arc today's material should advance; which loop to close or sharpen; which motif to bring back (running gag). It must connect to a REAL beat — never invent events. beat_kind hints at the dramaturgy: 🌀 twist / 💥 fail = cliffhanger; 🏆 milestone = payoff; 📡 external = fuel for the season's intrigue.
  3. Output 1-2 options: "arc X → sharpen (loop '…'), bring back the motif '…', anchored on beat-YYYY-MM-DD-slug". After that /episode writes the episode.

intake — process the beats inbox (sensor candidates → canon)

At night this is done by the routine canon-writer-nightly (01:00); here it is the same thing on demand. python $IMPORTS_ROOT/content-factory\canon_intake.py list → judge fitness and the ARC (only from the live dictionary!) → ... accept <beat_id> --arcs arc-a,arc-b or ... reject <beat_id> --why "...". The script itself: validates arc_id against the arc files, sets the derived story_day, moves it into beats\, appends the back-link into the arc, cleans _README. Never move beats by hand. canon_intake.py verify — the detector of half-finished intake (a beat sits in beats\ while the arc does not reference it: a break, or a race between two writers). Runs at the end of the nightly pass; exit 1 = fix it.

board — the season board (derived, not handwritten)

python $IMPORTS_ROOT/content-factory\show_canon_sync.py board (before→after) / board --apply (write it). season-01.active_arcs / open_loops are rebuilt from the status: fields in the arc/loop files. Rule: the board is never edited by hand — otherwise the copy falls behind the truth (the BOARD-DRIFT incident 07-27: an arc lived in the files for 16 days while the board did not show it). show_canon_sync.py arcs prints the live arc_id dictionary — the contract for the sensors.

Updating the state = EDITING THE CANON (with explicit actions, not silently)

Everything that used to be written into json is now an edit of canon cards, from the templates:

  • a new line → arcs\arc-<slug>.md (modelled on the existing ones); escalation/closure → edit the arc's current_state/status;
  • open/close a cliffhanger → loops\loop-<slug>.md (status: open/closed + current_best_answer);
  • link an episode → append consequences to the anchoring beat ("episode

recap --arc <id> — "previously on"

The arc's beats (beats[] from the arc card, by occurred_on) + its open loops → a 1-2 line recap for the new episode.

check — the weekly health scorecard of the series

A deterministic engine (0 tokens): python $IMPORTS_ROOT/content-factory\show_canon_check.py check (+ status for context). Axes: (1) Continuity — the season question is set + ≥3 beats happened; (2) Change — the share of active arcs that have beats; (3) Loops — more than 4 open = a jam, more than 2 in one arc = overload; (4) Trust (manual, receipts; the script reports the share of fresh beats carrying evidence_refs). Plus board drift (season-01.open_loops vs the real loop files) and a jammed beats inbox. Exit 0=OK · 1=FLAGS (a diagnosis) · 2=could not measure. The judgement verdict belongs to the top model / the owner. Weekly (routine reality-show-health-weekly) or on demand.

HOW IT WIRES INTO THE PIPELINE

An event → a BEAT into the canon (template beats\_TEMPLATE-beat.md) → /reality-show next (pick the move) → /episode writes the episode → publication through the gates → /reality-show records the consequences IN THE CANON → canon_render.py refreshes the public registries.

BOUNDARIES

  • Read-only by default (status/recap/next/check); canon edits happen through explicit, announced actions.
  • The owner's voice = the top-tier model (if the session runs a weaker one, delegate the next judgement to a subagent with model:'opus').
  • Draft-first outbound: /episode publishes on a "+". Tier-2 outbound → pause and ask.
  • Privacy: secrets OUT; respect the reveal axis (never burn live_hold/spoiler_until in outbound recaps). Serialization comes from REAL beats, never from invention.
  • AK-47: one SKILL.md, no separate store any more (the canon IS the store). Canon laws: decision-single-canon-story-state + _SHOW-CANON.md.

Like this skill? It is part of second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.

Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).

Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.

Files (second-brain-starter-kit)
  • SKILL.md 7.1 KB
    ---
    name: reality-show
    description: >-
      Keep serialized build-in-public content continuous: read one canon file of season, arcs, beats
      and loops and say which narrative beat to advance next, so episodes chain into a series. Pairs
      with the episode adapter, which does the writing. Triggers: "/reality-show", "season state",
      "what arcs are open".
    license: MIT
    ---
    
    REALITY-SHOW v2 — the season continuity layer on top of a SINGLE CANON
    
    GOAL: make the stream of episodes read like a SERIES. `/episode` writes one episode with dramaturgy (palette [[style-reality-show]]); THIS skill remembers the SEASON and suggests the next narrative move.
    
    **SOURCE OF TRUTH = the canon:** `$OBSIDIAN_VAULT/04-Projects\show-canon\`
    (season-01.md · arcs\ · beats\ · loops\ · rules\; the map is `_SHOW-CANON.md`).
    ⛔ **season-state.json is FROZEN as of 2026-07-10** (`_imports\content-factory\season-state.json` + season_state.py) — it was the third diverging spine after THREADS/SHOW-STATE; decision `decision-single-canon-story-state`. Do NOT read it as truth, do NOT update it.
    
    ## MODES
    
    ### `status` (default, read-only, 0 LLM tokens)
    Read `season-01.md` (the season question, active arcs, open loops, the board) + the headers of `arcs\*.md` (status, current_state) + `loops\*.md` (open?) → a one-line digest for the owner: the season + arcs with their stakes + dangling cliffhanger loops + story_day. Nothing is changed.
    
    ### `next` — propose the beat for the next episode (authorial model, cheap)
    1. Read: season-01.md + open loops + fresh beats in `beats\` (happened, latest by occurred_on) + the 📝 inbox `_imports\content-factory\triage\posts.md` (fresh post material). Do NOT read the whole corpus.
    2. Judgement (top-tier model): which OPEN arc today's material should advance; which loop to close or sharpen; which motif to bring back (running gag). It must connect to a REAL beat — never invent events. `beat_kind` hints at the dramaturgy: 🌀 twist / 💥 fail = cliffhanger; 🏆 milestone = payoff; 📡 external = fuel for the season's intrigue.
    3. Output 1-2 options: "arc X → sharpen (loop '…'), bring back the motif '…', anchored on beat-YYYY-MM-DD-slug". After that `/episode` writes the episode.
    
    ### `intake` — process the beats inbox (sensor candidates → canon)
    At night this is done by the routine `canon-writer-nightly` (01:00); here it is the same thing on demand.
    `python $IMPORTS_ROOT/content-factory\canon_intake.py list` → judge fitness and the ARC (only from the live dictionary!) → `... accept <beat_id> --arcs arc-a,arc-b` or `... reject <beat_id> --why "..."`.
    The script itself: validates arc_id against the arc files, sets the derived `story_day`, moves it into `beats\`, appends the back-link into the arc, cleans `_README`. Never move beats by hand.
    `canon_intake.py verify` — the detector of half-finished intake (a beat sits in `beats\` while the arc does not reference it: a break, or a race between two writers). Runs at the end of the nightly pass; exit 1 = fix it.
    
    ### `board` — the season board (derived, not handwritten)
    `python $IMPORTS_ROOT/content-factory\show_canon_sync.py board` (before→after) / `board --apply` (write it).
    `season-01.active_arcs` / `open_loops` are rebuilt from the `status:` fields in the arc/loop files. Rule: **the board is never edited by hand** — otherwise the copy falls behind the truth (the BOARD-DRIFT incident 07-27: an arc lived in the files for 16 days while the board did not show it). `show_canon_sync.py arcs` prints the live arc_id dictionary — the contract for the sensors.
    
    ### Updating the state = EDITING THE CANON (with explicit actions, not silently)
    Everything that used to be written into json is now an edit of canon cards, from the templates:
    - a new line → `arcs\arc-<slug>.md` (modelled on the existing ones); escalation/closure → edit the arc's `current_state`/`status`;
    - open/close a cliffhanger → `loops\loop-<slug>.md` (status: open/closed + current_best_answer);
    - link an episode → append consequences to the anchoring beat ("episode <slug> published") and to the arc's beats[].
    After the edits: `python $IMPORTS_ROOT/content-factory\canon_render.py` (the GitHub registries update themselves; the linter will complain about broken fields).
    
    ### `recap --arc <id>` — "previously on"
    The arc's beats (beats[] from the arc card, by occurred_on) + its open loops → a 1-2 line recap for the new episode.
    
    ### `check` — the weekly health scorecard of the series
    A deterministic engine (0 tokens): `python $IMPORTS_ROOT/content-factory\show_canon_check.py check` (+ `status` for context). Axes: (1) Continuity — the season question is set + ≥3 beats happened; (2) Change — the share of active arcs that have beats; (3) Loops — more than 4 open = a jam, more than 2 in one arc = overload; (4) Trust (manual, receipts; the script reports the share of fresh beats carrying evidence_refs). Plus board drift (season-01.open_loops vs the real loop files) and a jammed beats inbox. Exit 0=OK · 1=FLAGS (a diagnosis) · 2=could not measure. The judgement verdict belongs to the top model / the owner. Weekly (routine `reality-show-health-weekly`) or on demand.
    
    ## HOW IT WIRES INTO THE PIPELINE
    An event → a BEAT into the canon (template `beats\_TEMPLATE-beat.md`) → `/reality-show next` (pick the move) → `/episode` writes the episode → publication through the gates → `/reality-show` records the consequences IN THE CANON → canon_render.py refreshes the public registries.
    
    ## BOUNDARIES
    - Read-only by default (`status`/`recap`/`next`/`check`); canon edits happen through explicit, announced actions.
    - The owner's voice = the top-tier model (if the session runs a weaker one, delegate the `next` judgement to a subagent with `model:'opus'`).
    - Draft-first outbound: `/episode` publishes on a "+". Tier-2 outbound → pause and ask.
    - Privacy: secrets OUT; respect the `reveal` axis (never burn live_hold/spoiler_until in outbound recaps). Serialization comes from REAL beats, never from invention.
    - AK-47: one SKILL.md, no separate store any more (the canon IS the store). Canon laws: `decision-single-canon-story-state` + `_SHOW-CANON.md`.
    
    <!--kit-footer-->
    
    ---
    
    **Like this skill?** It is part of [second-brain-starter-kit](https://github.com/tonydzi/second-brain-starter-kit): the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with `npx skills add tonydzi/second-brain-starter-kit`. Everything is open source and free, so take what you need.
    
    Flagships worth a look on their own: [secondop-panel](https://github.com/tonydzi/secondop-panel) (a second opinion from a panel of external models), [claude-memory-tidy](https://github.com/tonydzi/claude-memory-tidy) (stop your agent's memory from rotting), [telegram-mcp-kit](https://github.com/tonydzi/telegram-mcp-kit) (your own Telegram over MCP in about 15 minutes).
    
    Author: **Anton Dziatkovskii**, Palo Alto AI Research Lab. Telegram [@tonydzi](https://t.me/tonydzi) - WhatsApp [+1 341 222 9178](https://wa.me/13412229178) - X [@Tony_Stef_](https://x.com/Tony_Stef_)
    
    **Engineers: want to test-drive this setup?** Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.