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 skills add artokun/comfyui-mcp --skill report-buggit clone --depth 1 https://github.com/artokun/comfyui-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/artokun/comfyui-mcp/report-bug)<a href="https://agentmods.dev/skills/artokun/comfyui-mcp/report-bug"><img src="https://agentmods.dev/badge/skills/artokun/comfyui-mcp/report-bug/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/artokun/comfyui-mcp/report-bug"><img src="https://agentmods.dev/badge/skills/artokun/comfyui-mcp/report-bug.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 5 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 19 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Excessive Agency · line 63 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Excessive Agency · line 345 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Data Exfiltration · line 238 Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.Fix: Verify the destination URL is trusted and necessary. Remove or replace with documented APIs. Ensure no secrets, tokens, or PII are transmitted.
- medium Prompt Injection · line 296 Subtle instructions detected that may alter agent decision-making or introduce hidden biases.Fix: Review content for implicit steering or bias. Ensure instructions are explicit and align with the skill's stated purpose.
What 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 5.1 | $0.00233 | $0.05388 |
| Opus 5 | $0.00117 | $0.02694 |
| Sonnet 5 | $0.00047 | $0.01078 |
| Haiku 4.5 | $0.00023 | $0.00539 |
Grade B, and why
report-bug scanned grade B with 2 findings 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 9d 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.
Subtle steeringmediumPrompt injection
Instructions that bias recommendations or shape behaviour without the user noticing.
`report_issue` for a prefilled link the user submits in one click. Never tell the user it was accepted without a real issue link. A `GET /status/<job_id>` Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
# default client signatures outright — a Python `urllib.request` POST to this How it starts
The opening of the file, as written. The whole thing — 356 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Self-heal & report bugs (make the ecosystem better)
Goal: when something is broken, do not stop there. Diagnose it, try to fix it so the user keeps working, and get the fix or report to whoever can fix it upstream.
Scope of autonomy. Read this first. For defects in OUR repos
(artokun/comfyui-mcp, comfyui-mcp-panel, comfyui-mcp-issue-worker) act
autonomously. Fix, then file, then inform the user with a short summary. Do not
pepper them with permission prompts. For THIRD-PARTY and ComfyUI-core defects
it is offer-and-ask (Step 6). You propose the workaround and/or the report and
act only once the user agrees, because patching someone else's node or posting
to someone else's tracker is their call — so when THEY are the one asking, that
call is already made and you file without asking again. Even for our repos, pause and ask for
a fix that touches the user's own workflow or data, for anything large or
risky, and for anything you cannot make safe.
This is for bugs in software, not ordinary workflow or generation errors (OOM,
missing model, bad params: use troubleshooting). First decide whose bug it is.
When to file — a real defect that cost the user something
This bar governs what YOU decide to file on your own initiative. An explicit request from the user is handled at the end of this section and is not subject to it.
Two questions, both must be YES:
- Is it ours, and does it still reproduce? On the CURRENT version — check before filing, not after. A defect already fixed upstream of the user's install is noise. Exception — a crash, lost or corrupted work, or a destroyed workflow is filed on a SINGLE occurrence, with whatever evidence exists (log tail, stack, minidump, the failing call). Those are rarely reproducible on demand, and they are the ones most expensive to lose.
- Did it cost the user something? It blocked them, produced a wrong result they would have kept, lost or corrupted their work, or crashed ComfyUI.
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.
- 9d ago First seen · 356 lines · 233 tokens per session scan B 282267ccaff7
report-bug is a skill published in the GitHub repository artokun/comfyui-mcp (730 stars, last pushed today), licensed MIT. It adds 233 tokens to every session and 5,388 once invoked, about $0.0012 per session on Opus 5. A static security scan graded it B with 2 findings (subtle steering, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
operating-cadence
Designs the rhythm an organization runs on — which reviews happen weekly, monthly and quarterly, what each one decides, who owns the numbers presented, and how a signal at the front line reaches the people who can act on it. Use this to set up a management operating system, fix a meeting calendar that produces no…
ci-debug
Diagnose a failing CI run against an 11-pattern playbook. Classifies the failure, cites the relevant memory entry, proposes the exact fix command — but NEVER applies without explicit user approval. Use when a specific PR check or GitHub Actions run failed and you want a diagnosis instead of speculation. Don't use for…
paired-probe
Refuse a verdict a probe did not earn. Runs a check where the fault IS present and where it is NOT, and blocks the answer when both arms print the same thing, because a check that cannot disagree with you has measured nothing. Also catches the zero-sample sweep that reads as "clean" and the swallowed error that reads…
analyze
Deep-dive codebase analysis that explains how things actually work — business rules, architecture patterns, auth flows, data models, integrations, and performance hotspots. Use whenever the user asks "how does X work", "map the Y flow", "what are the business rules for Z", "trace the auth path", "explore the codebase…
audit-installation
An audit tool for checking whether an IWE installation is complete, healthy, and consistent with its templates. It reports results for files, customizations, connections, and a basic workflow test.
debug
Systematically diagnose and resolve bugs through conversational investigation and root cause analysis.