Updated 3d ago
by lehoanganhtai13
archmancer
Eight skills for coding agents building systems where an agent selects model and action steps toward an outcome: map the architecture, choose the smallest design that meets the requirements, and trace failures across component boundaries.
22 GitHub stars (22)
+22 in 3 days
| Period | New stars |
|---|---|
| Nov 22 | +1 |
- 8 skills (8 skills found in its SKILL.md files)
- 2 forks (2 forks)
- 3d ago last update (Last updated 3d ago)
Install
Skills CLI
npx skills add lehoanganhtai13/archmancer
Asks which skills and agents to install. Just one: npx skills add lehoanganhtai13/archmancer --skill agent-configuration-and-delivery — add -g for every project.
Claude Code
git clone https://github.com/lehoanganhtai13/archmancercp -r "archmancer/skills/agent-configuration-and-delivery" ~/.claude/skills/
Copies the agent-configuration-and-delivery skill into ~/.claude/skills/ for every project (use .claude/skills/ for one project). Repeat for the other skills listed below.
Plugin
/plugin marketplace add lehoanganhtai13/archmancer
Run inside Claude Code if the repository ships a plugin marketplace, then pick plugins from /plugin.
Skills can run code on your machine — read the SKILL.md first.
What it does
In the author's own words, from each skill's SKILL.md.
-
agent-configuration-and-delivery
Use when model, prompt, provider or feature settings must reach running code: config validation, versioning, publishing, rollout and rollback, per-instance drift, settings that do not take effect after deploy, and proving which version and provider each instance actually serves.
skills/agent-configuration-and-delivery/ on GitHub (opens in a new tab)SKILL.md -
agent-context-and-state
Use when deciding what an LLM sees on each call: context assembly, history, summaries and compaction, prompt caching, progress or memory across sessions, context-window limits of local or hosted models, and diagnosing lost, repeated or contradictory context.
skills/agent-context-and-state/ on GitHub (opens in a new tab)SKILL.md -
agent-diagnostics-evaluation-and-resource-control
Use when you need evidence about LLM runs: instrumentation, traces, token and cost accounting, latency and budget limits, fixed before/after comparisons and regression rules, evaluation sets, and investigating slow, wrong, drifting or unexplained-cost behaviour.
skills/agent-diagnostics-evaluation-and-resource-control/ on GitHub (opens in a new tab)SKILL.md -
agent-evidence-and-knowledge
Use when an LLM answer must rest on data: files, SQL or spreadsheets, search APIs, a document corpus or RAG; computing numbers in code rather than in the model, keeping sources fresh after updates or corrections, citing versions, and telling 'no data' apart from 'source unavailable'.
skills/agent-evidence-and-knowledge/ on GitHub (opens in a new tab)SKILL.md -
agent-harness-model-and-tools
Use when an LLM feature must choose between a fixed workflow, one agent or several, call tools or external APIs, talk to a model provider or gateway, or add plugins and adapters; also for wrong tool calls, malformed provider requests, model routing, shared rate limits and quota across replicas, retries and 429 handling, and cost-saving levers such as smaller models, caching or batching.
skills/agent-harness-model-and-tools/ on GitHub (opens in a new tab)SKILL.md -
agent-runtime-and-interaction
Use when an LLM feature runs as a request, job, cron, queue or long-running task: admission, scheduling, cancellation, durable recovery, streaming and reconnect, idempotency and duplicate work, stalled or half-finished runs, and how results reach users safely (old output stays live until the new one is validated).
skills/agent-runtime-and-interaction/ on GitHub (opens in a new tab)SKILL.md -
agent-system-engineering
Start here for any software that calls an LLM: chat or support assistants, bots, CLIs, batch jobs that generate text, RAG or document search, tool-using agents, multi-step model pipelines. Use it whenever you plan, design, debug, review or cost such a feature, even if the request never says agent: wrong, stale or invented model output, slow or expensive model calls, prompts and context, provider…
skills/agent-system-engineering/ on GitHub (opens in a new tab)SKILL.md -
agent-trust-and-privacy
Use when an LLM feature touches users, money, credentials or untrusted input: authorization and isolation, human approval, payment or write limits enforced outside the model, tool credentials, prompt injection through uploads, retrieval or memory, and protecting personal data.
skills/agent-trust-and-privacy/ on GitHub (opens in a new tab)SKILL.md
No skill matches that filter.
Review before you install
A skill is a set of instructions and scripts your agent will follow, and it can run code on your machine. Read the SKILL.md and any scripts first — this listing is not a security review.