LLMSkillHub
mirroredCopied from amirfish1/claude-command-center under MIT. amirfish1 has not claimed this page, and is not affiliated with LLM SkillHub.

amirfish1 / press-room

Use when one release needs to go out across several channels at once — spawns one writer sibling session per channel via CCC (changelog, LinkedIn, X thread, optional blog) off a single source pack, then collates the drafts.

0
SKILL.md
4.6 KB

Press-room turns one release into N channel drafts in parallel: you build a
single source pack, spawn one writer session per channel each constrained to its
format, and collate the returned drafts into per-channel files. It builds on the
spawn/report_to mechanics in the ccc-orchestration skill — read that first for
the full Spawn/Inject/Ask API.

Cost

1 spawned session per channel — default 3 (changelog, LinkedIn, X); the
optional blog post makes 4. Each spawn is a real billed session on the user's
kanban — state the count before spawning.

Not for a one-channel announcement — write that yourself, or use a-b-copy if you
want to compare two voices on it.

Setup

CCC_URL="$(cat ~/.claude/command-center/port.txt 2>/dev/null || echo "${CCC_URL:-http://127.0.0.1:8090}")"
curl -sf --max-time 2 "$CCC_URL/api/version" >/dev/null 2>&1 || CCC_URL="http://127.0.0.1:8090"
curl -sf --max-time 2 "$CCC_URL/api/version" >/dev/null 2>&1 || CCC_DOWN=1

Notes: (a) run every CCC curl with the network sandbox disabled — the Bash
sandbox blocks loopback and fails spuriously even when CCC is up. (b) URL-encode
repo_path in query strings (+ → %2B, space → %20). (c) Resolve your own
session id from $CLAUDE_SESSION_ID, else the newest *.jsonl under
~/.claude/projects/<slugified-cwd>/ (see ccc-orchestration).

Step 1 (you): build the source pack ONCE

Write a single source-of-truth file (e.g. /tmp/press-room-<version>.md) so the
channels can't drift apart or invent claims. Include:

  • The release diff or commit list — git log vPREV..vNEW --oneline plus the key
    diffs.

  • What user-visible pain each change removes.
  • Relevant links/URLs the posts may use.

Every writer reads this file; none of them re-derives the facts.

Step 2: spawn one writer per channel, in parallel

All with report_to set to your own session id, each pointing at the source-pack
file plus its channel constraints. Omit "model" and "reasoning_effort": the
server spawn defaults (Settings → Spawn defaults) apply. Pass either explicitly
to keep cost down, with "engine" alongside it, since both are validated per
engine (see ccc-orchestration § 2 for the per-engine effort ladders).

curl -s -X POST "$CCC_URL/api/sessions/spawn" -H "Content-Type: application/json" -d @- <<'JSON'
{
  "prompt": "Write the <CHANNEL> draft for this release. Read the source pack at /tmp/press-room-<version>.md — it is the ONLY source of facts; invent nothing. Channel constraints: <CONSTRAINTS>. Report the draft only, no commentary.",
  "repo_path": "/abs/path/to/repo",
  "report_to": "<your-session-id>"
}
JSON

Bake in the per-channel <CONSTRAINTS>:

  • CHANGELOG — Keep-a-Changelog voice, user-visible changes only, no
    marketing adjectives.

  • LINKEDIN — 150-250 words, hook in the first line, one claim + one proof,
    CTA link, no hashtag spam.

  • X / THREAD — first post stands alone, every post ≤280 chars, 3-6 posts,
    one idea per post.

  • BLOG (opt-in) — 500-900 words, problem → what changed → how to use it, code
    examples only from the source pack.

Waiting for the report(s)

After spawning, end your turn and tell the user which channels are being
drafted. Drafts arrive by injection (the report_to footer CCC appends). Never
poll or sleep-loop (see ccc-orchestration). If a draft never arrives, check
GET /api/sessions/spawned for whether that writer's session is still alive.

Step 3: collate

As drafts arrive, write each into release-comms/<version>/<channel>.md (one file
per channel) and present the full set to the user. Drafts only — nothing is
posted or published; publishing is a human decision.

Dry run

If the arguments contain dry-run, print the channel list and session count, the
source-pack path, and each channel's constraint block and prompt — then POST
nothing and write no comms files.

When CCC is down

If CCC_DOWN=1 or a curl returns connection refused (exit 7): do not pretend it
ran. Fall back to one Task-tool subagent per channel with the same prompts —
same output, less visibility — and tell the user CCC was unreachable.

What this is not

Not one session writing all channels — the voices bleed and the X thread reads
like the blog. Not a-b-copy (that compares two voices on ONE piece; this fans
one release out to MANY channels).