project-learningslisted

Use PROACTIVELY, without being asked, at the START of any multi-step task in a git repo, project folder, or vault (before reading files) and at the END of one (to reflect and save). Applies whenever the user asks to implement, fix, refactor, debug, or investigate something in a project, whenever a first attempt was wrong or the user corrected the agent, and whenever the agent rediscovers something a previous session should have known. Retrieves what earlier sessions learned about operating in the project and saves durable insights (locations, traps, conventions, decisions with rationale, workflows, open questions, what to do differently) to an opt-in, append-only InfraNodus graph named learn- plus the project name. Also on "what do we know about this repo?", "what did I learn?", "save learnings", "remember this about the project", "start keeping learnings", "/learnings". Load before calling get_project_learnings, add_project_learnings, or enable_project_learnings directly. Never writes without the user's expl
infranodus/skills · ★ 118 · Code & Development · score 76

Install: claude install-skill infranodus/skills

# Project Learnings An agent that works in a project keeps rediscovering the same things: which of nine files holds the model list, that the tests only exist as ad-hoc scripts, why a cache is per-process, which command actually deploys. This skill turns those rediscoveries into a **learnings graph** in the user's InfraNodus account — `learn-<project>` — that the next session reads before it starts and appends to when it finishes. It is a graph rather than a notes file because learnings link to the same entities (file paths, modules, concepts) as everything else in the project: "which parts of the codebase accumulate traps" is a graph query, and the gaps between what is known and what the code contains show where nobody has looked yet. ## Prerequisites The connected InfraNodus MCP server must expose these three tools (added after `infranodus-mcp-server` 1.7.4; hosted at https://mcp.infranodus.com or the npm package): - `get_project_learnings` — read (by prompt, by entity, overview, or list) - `add_project_learnings` — propose / write (dry run by default) - `enable_project_learnings` — create the graph, **only on the user's explicit request** If they are missing from the tool list, say so once ("the connected InfraNodus server doesn't expose project learnings; update it or connect https://mcp.infranodus.com") and continue the task without this skill. Do not emulate the feature with `memory_add_relations` — that would bypass the consent model. ## Consent — read before any