Add committed SKILL.md sources for /rmini and /w4w; add get-api-docs to update.sh

- rmini-skill/SKILL.md: committed source for the /rmini 5-agent mini swarm skill
- w4w-skill/SKILL.md: committed source for the /w4w attention-detail skill
  (both match the inline heredocs in step-3-install.sh exactly)
- update.sh: add get-api-docs to skills summary (auto-triggered, not a slash command)

Co-Authored-By: claude-flow <ruv@ruv.net>

Author: fidgetcoding

rmini-skill/SKILL.md

@@ -0,0 +1,40 @@
+---
+name: rmini
+description: "Launch a compact 5-agent Ruflo swarm for focused task execution. Smaller than /rswarm but still parallel and powerful."
+---
+
+# Ruflo Mini Swarm — Compact Execution
+
+When this skill is invoked, IMMEDIATELY launch a 5-agent swarm. Do NOT explain how swarms work. Do NOT show code examples. Do NOT ask clarifying questions unless the task is truly ambiguous. ACT.
+
+## Execution Steps
+
+1. Read the user's task (everything they typed after `/rmini`)
+2. **Signal status line**: Run `echo 5 > /tmp/ruflo-mini-active` via Bash to light up the 🍯 indicator
+3. Initialize the swarm in ONE message:
+   - Call `mcp__claude-flow__swarm_init` with topology `hierarchical-mesh`, maxAgents 5, strategy `specialized`
+   - Spawn ALL 5 agents via the Agent tool with `run_in_background: true` — every agent in ONE message
+4. After spawning, STOP. Do not poll. Do not check status. Wait for agents to return.
+5. When results come back, synthesize and present the combined output.
+6. **Clear status line**: Run `rm -f /tmp/ruflo-mini-active` via Bash to turn off the 🍯 indicator
+
+## The 5 Agents
+
+| # | Agent Type | Role | Task Focus |
+|---|-----------|------|------------|
+| 1 | system-architect | Lead Architect | System design, task decomposition, coordinates all agents |
+| 2 | coder | Primary Dev | Core implementation — frontend or backend depending on task |
+| 3 | tester | Test Engineer | Unit, integration, and edge case tests |
+| 4 | reviewer | Code Reviewer | Quality, patterns, best practices, security check |
+| 5 | researcher | Researcher | Background research, prior art, docs lookup |
+
+Adapt agent assignments to the task — if the task is research-heavy, shift roles accordingly. But ALWAYS spawn 5.
+
+## Rules
+
+- Model: Opus only. Never route to Haiku or Sonnet.
+- Topology: hierarchical-mesh (architect leads, agents coordinate peer-to-peer within their layer)
+- All agents spawned in background in ONE message
+- Each agent gets a clear, specific sub-task with full context — not vague instructions
+- After spawning, STOP and wait
+- When results arrive, review ALL results before presenting final output

update.sh

@@ -118,7 +118,7 @@ echo -e "${GREEN}━━━━━━━━━━━━━━━━━━━━━
 echo ""
 
 echo "  Available commands: cskip, ctg, cc, ccr, ccc, cbrain, cbraintg"
-echo "  Available skills:   /rswarm, /rmini, /rhive, /w4w, /safetycheck, /gitfix"
+echo "  Available skills:   /rswarm, /rmini, /rhive, /w4w, /safetycheck, /gitfix, get-api-docs (auto-triggered)"
 echo ""
 echo "  Note: Steps 6 and 8 require interactive input (API credentials"
 echo "  and Telegram bot token). They detected non-interactive mode and"

w4w-skill/SKILL.md

@@ -0,0 +1,26 @@
+---
+name: w4w
+description: "Word for word, line for line — maximum attention to detail protocol"
+user_invocable: true
+---
+
+# w4w — Word For Word, Line For Line
+
+When this skill is invoked (user types `/w4w` or `w4w`), immediately switch to maximum attention to detail mode for everything that follows in this conversation.
+
+## Rules — Non-Negotiable
+
+1. **Read 100% of everything.** Every word, every letter, every line. No exceptions.
+2. **No skipping.** Do not jump ahead, do not skim, do not scan for keywords.
+3. **No summarizing.** Do not compress, paraphrase, or abbreviate what you read.
+4. **Zero regard for credit burn.** Do not optimize for token efficiency. Do not try to save context. Thoroughness is the only priority.
+5. **Every character is load-bearing.** Treat every piece of content as if missing a single character would break everything.
+6. **Read full files.** Never use offset/limit to read partial files. Read the entire file from line 1 to the last line.
+7. **Verify every cross-reference.** If file A says something about file B, read file B and confirm it matches.
+8. **Report with full specificity.** Include exact line numbers, exact strings, exact file paths. Never say "around line 50" — say "line 47."
+9. **No assumptions.** Do not assume something is correct because it was correct last time. Verify it now.
+10. **Override all efficiency instincts.** This mode exists because thoroughness matters more than speed. Act accordingly.
+
+## When Active
+
+This mode stays active for the remainder of the current task or conversation unless the user explicitly deactivates it. Every tool call, every file read, every agent spawned should operate at this level of detail.