From complexity-awareness
Use when a coding task has already shown complexity markers in this session — your initial hypothesis about the system has been wrong at least once, test failures or runtime behavior have surprised you, the same fix has been re-attempted with variations, or you've realized the user's framing of the problem may itself be approximate. Do not use on first contact with unfamiliar code, do not use because a task touches many files, and do not use for well-scoped tasks with a clear right answer (typical bug fixes, isolated feature additions, well-defined refactors). The trigger is observed surprise, not novelty. The signal is "the system keeps surprising me" or "I'm not sure I'm even working on the right thing."
How this skill is triggered — by the user, by Claude, or both
Slash command
/complexity-awareness:complex-task-modeThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You're in this skill because the task shows signs of being complex rather than merely complicated. The working mode shifts from plan-execute to probe-sense-respond.
You're in this skill because the task shows signs of being complex rather than merely complicated. The working mode shifts from plan-execute to probe-sense-respond.
Complex coding terrain has tells. If two or more of these are true, treat it as complex:
If only one of these is true, you're probably in complicated terrain with one uncertain element. Handle the uncertain element specifically and proceed normally.
In complex terrain, your plan is provisional. You're not executing a plan; you're running a sequence of small probes, each of which teaches you something about the system.
A probe is:
After each probe, ask: did the system behave the way I expected? If yes, your model is holding; continue. If no, your model just moved — stop and re-orient before the next action.
The most common complex-task failure is acting on a stale mental model. You believed something three actions ago that is no longer true, and you've been compounding the error since.
Drift signals:
When you notice drift: stop. Don't fix it by doing more. Read the actual current state of the relevant code or system. State out loud (in your reasoning) what you now know that you didn't before. Then decide the next probe from the new model, not the old one.
Your edits change the territory. The codebase after your last change is not the codebase you were modeling two changes ago. Re-read after substantive changes; don't trust your post-edit prediction of file state.
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub albeorla/complexity-awareness --plugin complexity-awareness