agentleFS
tatsoft-llctatsoft-llc/FrameworX-Projects/.claude/skills/manage-context-on-large-solutions/SKILL.md

manage-context-on-large-solutions

Skill1 starsChanged 27 days agoApache-2.0View on GitHub

Keep a FrameworX build coherent when the solution is too large to hold in context at once. Use when a build spans hundreds of tags or many displays, when the session is long, when the model starts losing track of what was built, or when the user says "this is getting big", "manage the context", "build this in batches", "the solution is large", or "don't lose the thread". Provides the method for scoping by layer and area, browsing instead of dumping, verifying each batch, and leaving a resumable trail so a later turn or a fresh session can continue without re-deriving.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Browse the part you are changing: Use browse_object_model, runtime_search_uns, and get_objects on a named scope.

    line 25
  2. Distinctive:

    All UserTypes (small, do them together.

    line 36
  3. Very specific:

    Verify each batch before starting the next (get_objects round-trip, or a runtime check for live behavior).

    line 42
  4. Distinctive:

    When a build spans many turns, write down where you are, so a later turn or a fresh session can pick up without re-deriving:

    line 52
  5. Distinctive:

    Keep a short running checklist in the spec or a session note: which areas are built and verified, which are next, any decision made …

    line 55

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Manage context on large solutions

A production solution is large. Hundreds or thousands of tags, dozens of displays, many scripts. You cannot hold all of it in context, and trying to is how a build loses coherence: the model re-reads the same dump, forgets what it already changed, and starts contradicting earlier work. The fix is to work in scoped batches and keep a trail.

Scope, do not dump

  • Browse the part you are changing. Use browse_object_model, runtime_search_uns, and get_objects on a named scope. Do not pull the whole object model into context to change one area.
  • Work the spec and the asset tree, not the full export. The spec says what to build; the asset tree says where it goes. Those two are small. The full object dump is large and mostly irrelevant to the current batch.

Batch by layer, then by area

Build in the order of KnowledgeBase/Method/build-order.md, and within a layer, by area:

  1. All UserTypes (small, do them together; everything binds to them).
  2. The UNS and asset tree for one area. Verify. Then the next area.
  3. Alarms for one area. Verify. Then the next.
  4. Historian for the logged tags. Verify.
  5. Displays, one screen or one Symbol at a time. Verify each with a screenshot.

Verify each batch before starting the next (get_objects round-trip, or a runtime check for live behavior). A verified batch is one you can stop thinking about. An unverified batch is weight you carry into every later turn.

A single display of hundreds of elements is a batch of its own: write it through a JSON workspace rather than an inline payload (KnowledgeBase/Patterns/MCP-Display-Authoring-Recipes.md, "Large displays").

Leave a resumable trail

When a build spans many turns, write down where you are, so a later turn or a fresh session can pick up without re-deriving:

  • Keep a short running checklist in the spec or a session note: which areas are built and verified, which are next, any decision made mid-build.
  • Record decisions and non-obvious constraints in the site's Decisions/ folder (see KnowledgeBase/Method/Memory-Guidance.md).
  • Name the sandbox clearly (ai-test-<slug>) so it is obvious what it is on return.

Signs you are losing the thread, and the recovery

  • You are re-reading objects you already changed: you lost the trail. Stop, write the checklist, resume from it.
  • The model proposes something the spec already settled: re-read the relevant spec section, not the whole solution.
  • Context is genuinely full: summarize the build state into a session note and the spec checklist, then continue from that summary rather than the raw history.

The principle

A large build is many small, verified builds in a defined order, with a written trail between them. It is not one heroic pass. Aim high on the whole plant; start simple on one area; scale by adding the next area to a foundation that already holds.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.