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 mhawthorne/gza --skill gza-system-triagegit clone --depth 1 https://github.com/mhawthorne/gzaWrote 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/mhawthorne/gza/gza-system-triage)<a href="https://agentmods.dev/skills/mhawthorne/gza/gza-system-triage"><img src="https://agentmods.dev/badge/skills/mhawthorne/gza/gza-system-triage/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/mhawthorne/gza/gza-system-triage"><img src="https://agentmods.dev/badge/skills/mhawthorne/gza/gza-system-triage.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, 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 31 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.
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 | $0.00145 | $0.05137 |
| Opus 5 | $0.00072 | $0.02568 |
| Sonnet 5 | $0.00029 | $0.01027 |
| Haiku 4.5 | $0.00015 | $0.00514 |
Grade A, and why
gza-system-triage scanned grade A with 0 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 7d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 208 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Gza System Triage
Every stuck task is a failed auto-merge. This skill does three things, in order, for the pile watch leaves behind:
- Diagnose — bucket the stuck rows by failure class and find why each class is stuck.
- Unstick now — clear the existing stuck rows by their correct action: drop the moot/dead/stale, spawn the missing follow-up, or hand review/improve-loop rows to
/gza-task-fix. A stuck row left in place is debt that grows faster than anyone can service it. - Prevent — for each class, file the broadest
system-tagged change that would let this and its siblings auto-merge unattended next time, ranked by blast radius.
Diagnosis without unsticking just re-describes the pile; unsticking without prevention means the pile rebuilds tomorrow. Do all three.
It exists because the same conversation recurs daily: watch leaves a pile of ATTENTION / Needs attention rows, and each session re-derives the landscape from scratch and ships another narrow micro-fix that doesn't compound. This skill closes that loop: snapshot once, bucket by class, skip classes already being fixed, clear what's clearable now, and file the broadest prevention fix for the rest.
The per-task rescue (/gza-task-fix) and per-lineage drop/rebase classification (/gza-task-triage) are the workers this skill drives — it owns the orchestration (which row gets which action) and executes the safe clearing actions itself, delegating only the heavy per-unit rescue.
What this skill MUST NOT do
These guardrails are load-bearing.
- Do not merge, retry, resume, or delete branches. If a row warrants one of those, say so and stop. (Note: dropping a task via
gza set-status <id> droppedis not deleting a branch — the branch stays as history — and is the sanctioned way to clear a moot/dead/stale row. Spawning a follow-up viagza add --based-on/gza implementand handing a row to/gza-task-fixare likewise allowed. These are the unstick actions; merge/retry/resume are not.) - Do not edit code. A single unit that needs a code fix is a
/gza-task-fixhandoff or a filed task, not an inline edit. - Confirm before bulk or destructive unstick actions in manual mode. Drops are reversible (
set-status ... pending) but bulk drops still need an explicit go-ahead; never mass-drop on a terse instruction without confirming scope. - Do not file untagged tasks. Every filed task gets
systemplus the active recovery tag — an untagged task is a silent orphan watch never runs. - Do not file anything in manual mode without confirmation.
- Do not propose another narrow guard for a class that already recurred after a fix landed. That is the signal the previous fix missed the cause layer — escalate to re-diagnose deeper (see
specs/behavior/systemic-fix-triage.md), never patch the symptom again. - Do not rescue a task without accounting for its auto-merge fix, and never re-file a fix that already landed. Every
/gza-task-fixrescue is a failed auto-merge — run the pairing-with-verification gate (Step 7) before filing any prevention. If the fix that would have auto-merged the task already merged yet the task still got stuck, file a cause-layer escalation, never another copy of that fix. A prevention that clears only the one rescued row is a symptom patch, not a systemic fix.
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.
- 7d ago Changed 9916438bf86d
- 11d ago First seen · 208 lines · 145 tokens per session scan A 63ab7cb27892
gza-system-triage is a skill published in the GitHub repository mhawthorne/gza (12 stars, last pushed today), licensed MIT. It adds 145 tokens to every session and 5,137 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
sentry-fix-issues
Find and fix issues from Sentry using MCP. Use when asked to fix Sentry errors, debug production issues, investigate exceptions, or resolve bugs reported in Sentry. Methodically analyzes stack traces, breadcrumbs, traces, and context to identify root causes.
audit
Run technical quality checks across accessibility, performance, theming, responsive design, and anti-patterns. Generates a scored report with P0-P3 severity ratings and actionable plan. Use when the user wants an accessibility check, performance audit, or technical quality review.
debugging-executions
Debug failed or wrong-output workflow executions using executions tools. Load when the user reports execution failures, unexpected node output, empty parameter values after a successful run, or a node showing a red or failed expression error.
n8n-docs-assistant
Answers n8n product, setup, credential, node, hosting, API, and usage questions from current n8n docs. Use when the user asks how to configure, set up, troubleshoot, or understand n8n behavior, especially credential setup questions opened from the credential modal.
dorodango
Polishes working code through successive quality passes in fresh subagents. Use after tests pass when code needs multi-dimension refinement before release.
sentry-sdk-upgrade
Upgrade the Sentry JavaScript SDK across major versions. Use when asked to upgrade Sentry, migrate to a newer version, fix deprecated Sentry APIs, or resolve breaking changes after a Sentry version bump.