CRITICAL. MUST activate at 65%+ context usage, 2h / 15+ file / 350+ line scope, or large-file restructuring. Proactive planning, reorganization, stale-content cleanup.
SkillScore75Maintained
Automatic
◆AI Engineer pack
Last verified 8/27/2026
What it changes
Passed 17 of 23 test cases with the skill, and 10 of 23 without it.
Without skill42%
With skill75%
Checks it passed that the model alone missed
Formats tasks inside a tasks XML tag
Includes summary hooks for long output
Formats task items inside a tasks XML tag
and 15 more in the full report
11 fixed7 passed either way3 still failing3 broke
Pass-rate lift
+33 pp
Tokens
-21%
Agent turns
0%
Response time
-32%
Model gemini-3.6-flashMeasured 8/27/2026v318 of 24 cases passed with skill · 23 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 11
Prompt
During an interactive debugging session on the authentication module (auth_service.ts), context utilization monitor reports 77% context window usage. What immediate warning action is required?
case-05 · gemini-3.6-flash · 8/27/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md15 linessingle file
<self_organization>
Plan proactively; run everything as todo tasks per always-on <tasks>, incl. subagent dispatch/orchestration.
Explicit plan items for: restructuring large in-scope files (~500+ lines / 10K+ size) · cleanup of stale/outdated/redundant content.
At 65% context → output "WARNING! High context consumption, consider using new session!"; at 75% → output "CRITICAL! Context consumption is very high, you must start a new session!".
Over 2h / 15+ files / 350+ line spec → propose scope reduction; user may override.
Max ~2 pages per review pass; TLDR/summary hooks for long outputs.
Announce self-organization intent in advance — keep user in loop before restructuring files, splitting scope, reducing output, or starting a new session.
Output overflow → write in batches, section-by-section.