agentsclimarketplace

Context engineering

Skill BuilderCed/agent-skills/skills/meta/context-engineering

31 cross-platform AI agent skills for regulated industries & underserved markets. EU compliance (AI Act, NIS2, DORA, GDPR), French professional (accounting, tax, notary, real estate), security audit, agent evaluation, Africa mobile money, offline-first.

Install
npx -y skills add BuilderCed/agent-skills --skill context-engineering

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Manage AI agent context effectively — what to include, what to exclude, compression strategies, and context hierarchy for optimal performance.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.7 KB, as published. Nobody here has run it

Context Engineering

Based on ETH Zurich research: overly detailed instructions reduce task success by 3%, increase token cost by 20%, and add 2-4 reasoning steps.

When to Use

  • Writing SKILL.md, AGENTS.md, or system prompts
  • Debugging poor agent performance
  • Optimizing token costs
  • Designing multi-agent workflows
  • Reducing context window pressure

Context Hierarchy (5 Levels)

Most persistent → most transient:

LevelContentPersistenceExample
1. RulesProject-wide standardsAlways loadedCLAUDE.md, AGENTS.md
2. SpecFeature/session scopePer featurePRD, architecture docs
3. SourcePer taskPer taskRelevant source files
4. ErrorsPer iterationPer attemptTest failures, stack traces
5. HistoryAccumulatesSessionConversation history

Principle: Levels 1-2 are curated (high leverage). Levels 3-5 are per-call (keep minimal).

What to Include

Include ONLY what the agent cannot discover independently:

  • Non-obvious conventions ("we use snake_case for DB columns")
  • Project-specific constraints ("never modify the auth module")
  • Architectural decisions not in code ("we chose Drizzle over Prisma because...")
  • External dependencies not discoverable ("deploy via internal CI, not GitHub Actions")

What NOT to Include

The agent can discover these itself — including them wastes tokens:

  • Tech stack (visible in package.json / requirements.txt)
  • File structure (visible via ls / find)
  • Key files (visible via search)
  • Build commands (visible in scripts / Makefile)
  • Standard patterns (the model already knows React, Express, etc.)

Sizing Guidelines

Context TypeMax SizeRationale
AGENTS.md500-1000 tokensETH Zurich: more = worse
SKILL.md (core)1000-2500 tokensBalance detail vs overhead
references/ per skill500-1000 tokensSupport data, not duplicate
System prompt total< 5K tokensBeyond this: diminishing returns

Compression Strategies

  1. Remove examples when the pattern is clear — one example > three redundant ones
  2. Use tables over prose — 50% fewer tokens for structured info
  3. Remove "obvious" instructions — "write clean code" is noise
  4. Use references for static data — move schemas/checklists to files
  5. Lazy-load context — only load what's needed for current task

Anti-Patterns

Anti-PatternProblemFix
"Always be thorough"Forces effort=high, +35% tokensRemove — model handles this
"Think step by step"Redundant with adaptive thinkingRemove on modern models
Repeating the same rule 3xToken waste, no benefitState once, clearly
Including full API docsContext overflowLink to docs, summarize key parts
"You are a helpful assistant"Generic, no valueUse specific task context

What This Skill Does NOT Do

  • Does not manage conversation memory (different problem)
  • Does not optimize the model itself (skill ≠ fine-tuning)
  • Does not handle multi-agent coordination (orchestration concern)

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.