agentleFS
ralphkraussralphkrauss/agent-orchestrator/.claude/skills/load-context/SKILL.md

load-context

Skill4 starsChanged 6 months agoMITView on GitHub

Load existing context files into the conversation for a feature branch. Use when user says "load context", "read context", "get context", or wants to resume work on a branch where /build-context was already run.

We split this skill into the rules it gives an agent

8 rules. 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    This is the complement to /build-context: Use /build-context to create context files.

    line 12
  2. Distinctive:

    If no context files found:

    line 20
  3. Distinctive:

    Read files in this order for optimal context placement:

    line 26
  4. Very specific:

    If the tiered files don't exist (legacy single-file format), read whatever .md files are present.

    line 32
  5. Very specific:

    api-context.md: check generated_at — this is cached and rarely stale

    line 38
  6. Very specific:

    project-context.md: check generated_at — if older than 1 day, warn the user:

    line 39
  7. Distinctive:

    Verify each file exists (skip missing with a note)

    line 50
  8. Distinctive:

    Read all existing critical and important files using parallel agents if there are many files (>10)

    line 51

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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
<!-- Generated from .agents/ by scripts/sync-ai-workspace.mjs. Do not edit directly. -->

Load Context

Load pre-built context files into the current conversation so you can immediately proceed with planning or implementation — no research phase needed.

This is the complement to /build-context. Use /build-context to create context files. Use /load-context to reload them in a new conversation. Use /update-context to update stale project context without re-extracting API docs.

Instructions

Step 1: Detect Branch and Find Context Files

  1. Get the current git branch: git branch --show-current
  2. Glob for context files: contexts/{branch-name}/**/*.md
  3. If no context files found:
    • Tell the user: "No context files found for branch {branch-name}. Run /build-context first."
    • Stop here.

Step 2: Read Context Files (tier-aware)

Read files in this order for optimal context placement:

  1. context.md (Tier 3 — combined summary) — read first for orientation
  2. api-context.md (Tier 1 — cached API reference) — read second for provider knowledge
  3. project-context.md (Tier 2 — dynamic project state) — read third for current state

If the tiered files don't exist (legacy single-file format), read whatever .md files are present.

Step 3: Check Freshness

Parse the YAML front matter from each tier file:

  • api-context.md: check generated_at — this is cached and rarely stale
  • project-context.md: check generated_at — if older than 1 day, warn the user:
    ⚠ Project context was built on {date}. Plans, reviews, and code may have changed.
    Run `/update-context` to update it (~1 min) or `/build-context` for a full rebuild.

Step 4: Read Must-Read Files

  1. From the context files, find the Must-Read Files section
  2. Extract file paths from Critical and Important tables (skip Reference)
  3. Deduplicate across all context files
  4. Verify each file exists (skip missing with a note)
  5. Read all existing critical and important files using parallel agents if there are many files (>10)

Step 5: Summary

Context loaded for branch `{branch-name}`.

Tiers:
  - API context: {generated_at} {(N days old)}
  - Project context: {generated_at} {(N days old)} {⚠ stale if > 1 day}

Context files read: {N}
Must-read files read: {N} critical, {N} important ({N} skipped — not found)

Critical path:
  1. {from context file}
  2. ...

Plan status:
  - Completed: {summary}
  - In progress: {summary}

Reference files available (not loaded): {N}

Ready for the next task.

Critical Rules

  • Never modify context files — this skill is read-only
  • Read in tier order — context.md first (orientation), then api-context.md (reference), then project-context.md (current state)
  • Skip reference-priority files — only read critical and important; mention reference files are available
  • Warn on stale project context — if project-context.md is older than 1 day, suggest /update-context
  • Deduplicate across files — if multiple files list the same must-read path, read it once
  • Fail fast if no context exists — suggest /build-context

Reference

  • .agents/skills/build-context/SKILL.md — creates context files (tiered)
  • .agents/skills/update-context/SKILL.md — lightweight update of project context

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.

Reviews can't be read right now.

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.