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-task-improvegit 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-task-improve)<a href="https://agentmods.dev/skills/mhawthorne/gza/gza-task-improve"><img src="https://agentmods.dev/badge/skills/mhawthorne/gza/gza-task-improve/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-task-improve"><img src="https://agentmods.dev/badge/skills/mhawthorne/gza/gza-task-improve.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00041 | $0.02854 |
| Opus 5 | $0.00020 | $0.01427 |
| Sonnet 5 | $0.00008 | $0.00571 |
| Haiku 4.5 | $0.00004 | $0.00285 |
Grade A, and why
gza-task-improve 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 today.
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 — 261 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Improve Gza Task Inline
Address feedback for a gza task directly in the current conversation. Feedback can come from two sources: a completed review (Must-Fix items plus Suggestions) and/or unresolved feedback task comments attached to the implementation. If no review exists but unresolved feedback comments do, improve still runs — comments alone are a valid feedback source. Other comment kinds such as review_scope remain visible to operators but are not improve-actionable and must not be resolved by this workflow. This skill is useful when a task has reached max review/improve cycles and needs human-guided fixes, or when you want to interactively resolve feedback.
Process
Step 0: Capture the starting checkout
Before touching task state, capture where the user started:
git symbolic-ref --quiet --short HEAD || git rev-parse --short HEAD
Save this as <START_CHECKOUT>. You may switch to the implementation branch to make changes, but before finishing you must return the user to <START_CHECKOUT>. If <START_CHECKOUT> is a detached HEAD, restore it with git checkout --detach <START_CHECKOUT>.
Step 1: Get task ID and find the feedback (review and/or unresolved comments)
The user should provide a full prefixed task ID (for example, gza-1234). If they provide a review task ID, resolve it to the implementation task. If no task ID is provided, ask the user.
Query the task, its most recent review, and any unresolved task comments:
uv run python -c "
import json, sys
from pathlib import Path
from gza.config import Config
from gza.db import SqliteTaskStore
config = Config.load(Path.cwd())
store = SqliteTaskStore.from_config(config)
task = store.get(<TASK_ID>)
if not task:
print('ERROR: Task not found', file=sys.stderr)
sys.exit(1)
# If the user gave us a review task, resolve to its parent implementation
impl_task = task
if task.task_type == 'review' and task.depends_on:
impl_task = store.get(task.depends_on)
elif task.task_type == 'improve' and task.based_on:
impl_task = store.get(task.based_on)
# Find latest review
assert impl_task.id is not None
reviews = store.get_reviews_for_task(impl_task.id)
latest_review = reviews[0] if reviews else None
# Find unresolved feedback comments (comments-only improve runs when no usable review exists)
unresolved_comments = store.get_comments(
impl_task.id,
unresolved_only=True,
kinds=('feedback',),
)
print(json.dumps({
'impl_task_id': impl_task.id,
'impl_task_type': impl_task.task_type,
'impl_branch': impl_task.branch,
'impl_prompt': impl_task.prompt,
'review_task_id': latest_review.id if latest_review else None,
'review_report_file': latest_review.report_file if latest_review else None,
'review_output': latest_review.output_content if latest_review else None,
'unresolved_comments': [
{'id': c.id, 'source': c.source, 'author': c.author, 'content': c.content, 'created_at': str(c.created_at)}
for c in unresolved_comments
],
'verify_command': config.verify_command,
'inner_verify_command': config.inner_verify_command,
}, default=str))
"
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.
- today Changed · +1 lines 518458c3f858
- 9d ago First seen · 260 lines · 41 tokens per session scan A 7765ed4fe063
gza-task-improve is a skill published in the GitHub repository mhawthorne/gza (12 stars, last pushed today), licensed MIT. It adds 41 tokens to every session and 2,854 once invoked, about $0.0002 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
ac-qa-gh-pr-review
Reviews pull requests through a bounded shared worker-wave review flow, consolidated reporting, and user-confirmed gh pr review actions.
Git Workflow
Use this skill when you want a clean Git history (easy reviews, easy rollbacks) and consistent PR hygiene—especially when an agent is helping you make changes across multiple files.
requesting-code-review
Use when requesting a focused review with standardized findings format (critical, important, informational) before commit or PR.
guardian
A guide for deciding how to organize commits, branches, merges, and pull requests. A pull request is a proposed code change submitted for review before it is merged.
template-proposal-review
Review proposal files in localdocs/localdocs/proposals/ and decide whether to apply them to the upstream python-project-template repository. Use when evaluating proposals generated by template-upstream, deciding which ones to accept or reject, and applying accepted ones to the template repo. Triggers on: "proposal…
review-implement-phase
Implements triaged review actions, commits focused fixes, and posts Done plus resolves threads. Use when the user wants only the implementation phase of the review-framework workflow.