wiki-queue-add

name	Wiki Queue Add
description	Enqueue wiki-worthy content (from a chat turn, spec write, PR merge, learning) for later agent-driven classification + ingest. Agents must justify with --reason.
version	1.0.0
category	knowledge
skillType	claude-skill
assignableRoles	["orchestrator","team-leader"]

Wiki Queue Add

Enqueue a candidate item for the LLM-wiki. This skill does not write to the vault — it adds the item to a queue. A separate wiki-process-queue run (later, batch or bookkeep-triggered) decides whether the item is actually worth saving and where in the vault it lands.

When to use

After a chat turn (yours OR a teammate's), pause briefly and ask:

• Did this turn produce content with lasting value?
  • a decision (pricing, scope, scheduling)
  • a fact about a person, customer, partner
  • a pattern that future-me would want to find
  • a new constraint, gotcha, or learning
• Was a fact mentioned that the team will need to look up later?

If YES, call wiki-queue-add. DO NOT call it for:

• Routine status checks ("standup", "?", "ok")
• Implementation details captured in code/PRs
• Conversation pleasantries
• Anything already in the wiki

Required arguments

Flag	Required	Purpose
--vault <path>	yes	Absolute path to the vault root.
--content <text>	yes	The text to capture.
--reason <text>	yes	One sentence: WHY is this wiki-worthy? Audit + helps the processor decide where it goes.
--source-ref <ref>	yes	Stable reference: chat msg id, slack thread, file path, WI id.
--source-type <type>	no	Default user_chat. Use slack_message, spec_file, pr_merge, record_learning, task_verified when appropriate.
--queued-by <sess>	no	Defaults to $CREWLY_SESSION_NAME or crewly-orc.

Examples

A pricing decision the team should remember:

bash execute.sh \
  --vault ~/Desktop/projects/crewly-projects/crewly/.crewly/wiki \
  --content "Anthropic SMB pilot signed at \$799/month, 12-month commit, 30-day onboarding clause" \
  --reason "first paid SMB customer + concrete pricing validation point + new contract pattern" \
  --source-ref "slack://D0AC7NF5N7L/1779509999.111"

A learning a worker just surfaced:

bash execute.sh \
  --vault ~/.crewly/teams/ad923b66-…/wiki \
  --content "Gemini text-embedding-004 returned 404 — deprecated. Falling back to keyword search." \
  --reason "infra gotcha — affects every team using vector search; future-me will hit this" \
  --source-ref "wi:6cf2d54d-…" \
  --source-type record_learning

Output

{
  "success": true,
  "item": {
    "id": "uuid",
    "status": "pending",
    "queuedAt": "2026-05-22T...",
    "vaultPath": "...",
    "content": "...",
    "reason": "...",
    ...
  }
}

The item sits in the queue until processed. Use wiki-queue-list to see pending items and wiki-process-queue to claim + classify + ingest.

Failure modes

• 400 reason is required — agents MUST justify wiki-worthiness; refusal protects future-me from low-signal noise
• 400 content/sourceRef empty — every queued item needs identifiable provenance
• 400 vaultPath must be absolute — relative paths are rejected; the vault root is part of the audit trail