Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/howar31/magi-workflow/helpnpx skills add howar31/magi-workflow --skill helpgit clone --depth 1 https://github.com/howar31/magi-workflowWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00101 | $0.02982 |
| Opus 5 | $0.00051 | $0.01491 |
| Sonnet 5 | $0.00020 | $0.00596 |
| Haiku 4.5 | $0.00010 | $0.00298 |
Grade C, and why
help scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
Try to read project state; never refuse. Help is unconditionally allowed, How it starts
The opening of the file, as written. The whole thing — 299 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/magi:help — quick reference & next-step hint
You are the coordinator. This skill is a read-only reference printer. Do not modify any files, do not call subagents, do not invoke external CLIs. Output goes to the user as plain markdown.
0. Resolve paths
PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-}"
[[ -z "$PLUGIN_ROOT" ]] && PLUGIN_ROOT="$(cd "$(dirname "$BASH_SOURCE[0]")/../.." 2>/dev/null && pwd)"
USER_CONFIG="$HOME/.config/magi-workflow/config.json"
Verify $PLUGIN_ROOT/skills is a directory. If not, the plugin is installed
incorrectly — print a one-line error pointing to the install instructions
and stop.
0.1. Output language
Read output_language from $USER_CONFIG if it exists; default zh-TW.
Missing config is not an error — /magi:help must work before
/magi:setup has ever been run, since "I don't know what to do" is exactly
when users reach for help.
LANG_PREF="zh-TW"
if [[ -f "$USER_CONFIG" ]]; then
LANG_PREF=$(jq -r '.output_language // "zh-TW"' "$USER_CONFIG" 2>/dev/null || echo "zh-TW")
fi
Section headers and ASCII flow diagrams stay English (matching README
conventions); descriptive prose follows LANG_PREF.
0.2. State preflight (fail-soft)
Try to read project state; never refuse. Help is unconditionally allowed, but the reading lets the overview append a "current state → suggested next" hint at the end.
STATE_JSON=""
if STATE_RAW=$(bash "$PLUGIN_ROOT/scripts/shared/detect-state.sh" 2>/dev/null); then
STATE_JSON="$STATE_RAW"
fi
If STATE_JSON is empty (not in a git repo, or detect-state errored), skip
the next-step hint section and emit the static overview only.
1. Argument parsing
/magi:help → overview mode
/magi:help <name> → detail mode for skills/<resolved>/SKILL.md
/magi:help --list → bare command list, no flow diagram or hint
Resolve <name>:
- Accept
plan,plan, or/magi:plan - Strip leading
/andmagi.prefix; the remainder must match a directory under$PLUGIN_ROOT/skills/magi.<remainder>/ - Unknown name → print
unknown command: <name>and the bare command list, then exit
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 299 lines · 101 tokens per session scan C 263909d3a107
help is a skill published in the GitHub repository howar31/magi-workflow (2 stars, last pushed 3d ago), licensed MIT. It adds 101 tokens to every session and 2,982 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it C with 1 finding (tells the agent never to refuse). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
model-compatibility
Model family compatibility matrix covering loaders, resolutions, samplers, CFG, VAE, ControlNet, and LoRA compatibility for SD 1.5, SDXL, Flux, SD3, and video models.
comfyui-node-registry
Authoring & publishing ComfyUI custom nodes to the Comfy Registry, covering node structure, pyproject.toml spec, comfy-cli publishing, and CI.
flux-txt2img
Build Flux txt2img workflows with Flux.1 Dev (SRPO), Flux 2 Klein 9B, Turbo LoRAs, FluxGuidance, and DualCLIPLoader patterns.
civitai
Discover Civitai models with the BUILT-IN downloadmodel action:"searchcivitai" and install/generate them locally. Find a checkpoint/LoRA/embedding on Civitai, download it into ComfyUI, and use its trigger words. Optionally pair the official Civitai MCP for community features (images browsing, posting, collections).
chief-strategy-officer
Owns where the business plays and how it wins over a multi-year horizon — portfolio choices, corporate development, strategic partnerships, and planning under uncertainty. Use this for a decision about which markets or businesses to be in, whether to build, buy, or partner, how to allocate capital across business…
performance-analysis
Measurement approaches, profiling patterns, bottleneck identification, and optimization guidance. Use when diagnosing performance issues, establishing baselines, identifying bottlenecks, or planning for scale. Always measure before optimizing.