context-engineering-collection

An agent skill by muratcankoylan, from muratcankoylan/Agent-Skills-for-Context-Engineering. Tags: api, debugging, developer-tools, strategy.

What it does

A comprehensive collection of Agent Skills for context engineering, harness engineering, multi-agent architectures, and production agent systems. Use when building, optimizing, evaluating, or debugging agent systems that require effective context management and reliable operating loops.

Install

With the skills CLI, which installs into Claude Code, Codex, Cursor and other agents:

npx skills add muratcankoylan/Agent-Skills-for-Context-Engineering --skill context-engineering-collection

Or copy the skill folder into Claude Code's skills directory by hand (~/.claude/skills for every project, or .claude/skills inside one):

git clone --depth 1 https://github.com/muratcankoylan/Agent-Skills-for-Context-Engineering
cp -r Agent-Skills-for-Context-Engineering ~/.claude/skills/context-engineering-collection

Safety box score

Grade F, high risk: 74 of 100 risk points, rated 2026-10-20. The grade rates what the skill and its scripts can reach on the machine of whoever installs it, across eight categories from shell execution to secrets access. How the score works.

This collection is mostly documentation and templates for building agent systems, but it also ships runnable validation, benchmark, and deployment tooling. The main risks are host persistence through launchd, destructive filesystem cleanup, and networked/API-backed workflows that can send data off machine.

Category levels, from 0 (none) to 3 (high)
Category	Level	Why
Code & shell execution	2 (moderate)	It ships and documents executable Python/Node tooling (`python ...`, `npm test`, and `subprocess.run(...)` in `researcher/scripts/run_benchmarks.py`), but the code is mostly local project automation rather than arbitrary downloaded code.
Network access	2 (moderate)	The repo can fetch remote sources and hit live model APIs (`--allow-fetch` in the research loop, plus OpenAI/Anthropic example configs), so data can leave the machine.
Filesystem reach	3 (high)	It writes queue/report files, can recursively remove a staging tree, and one runbook installs files under `~/Library/LaunchAgents/`, so it reaches outside the project and can alter user files.
Secrets & credentials	2 (moderate)	Example projects expect API keys in `.env` and use `dotenv`/`OPENAI_API_KEY`, but there is no sign of keychain or browser-session scraping.
Privilege & persistence	3 (high)	`researcher/runbooks/continuous-operation.md` installs launchd agents and bootstraps them into the current user agent domain, which changes OS-level startup behavior.
Instruction hijack surface	1 (low)	There is no clear prompt-injection bypass; the text is mostly governance and prompt-engineering guidance, though some templates discuss system-prompt shaping.
Irreversible actions	3 (high)	The collection explicitly covers autonomous research loops, long-running daemon jobs, PR preparation, and a GitHub Pages deploy-on-main workflow, so autonomous side effects are inherent to its purpose.
Transparency	2 (moderate)	Several referenced skill/script files were not readable in this review, so some behavior of the broader collection remains unverified.

Before installing

Read `researcher/runbooks/continuous-operation.md` and `researcher/orchestration/launchd/install.sh` before enabling the loop; it persists a user LaunchAgent.
Use a disposable staging directory when testing `researcher/scripts/validate_export.py`; it can delete the staging tree on failure.
Decide whether you want live model API usage from `examples/llm-as-judge-skills` before creating `.env` with `OPENAI_API_KEY`.
Confirm you are comfortable with automatic GitHub Pages deployment from `main` before merging prompt-lab changes.

Rated 2026-10-20 by an automated review of 72 files from the skill's repository.

Source

Repository: muratcankoylan/Agent-Skills-for-Context-Engineering (all skills from this repository)
Path: SKILL.md
Branch: main
Updated: 2026-09-19

tool-design — This skill should be used for the tool-interface layer of an agent system specifically: writing tool descriptions agents can route on.
bdi-mental-states — This skill should be used when modeling agent mental states with BDI concepts: beliefs, desires, intentions, RDF-to-belief transformations.
context-degradation — This skill should be used for diagnosing and mitigating context degradation: lost-in-middle failures, context poisoning, context clash, context confusion.
harness-engineering — This skill should be used when designing autonomous agent harnesses: research loops, evaluation scaffolds, locked and editable surfaces, durable logs.
memory-systems — This skill should be used for persistent semantic memory in agent systems: cross-session knowledge retention, entity tracking, temporal validity.
reasoning-trace-optimizer — Debug and optimize AI agents by analyzing reasoning traces, context degradation, tool confusion, instruction drift, repeated task failures.