Dev & Engineering github-issue-triagepull-request-reviewdraft-responseslanggraphfastapistreamlitgh-clihuman-in-the-loop

Maintainer Response (agent-service-toolkit)

Drafts issue and PR replies for the agent-service-toolkit repo in the maintainer's authentic voice — always as human-reviewed drafts, never posting on its own.

FollowSkills review · FSRS-1.0
Recommended
69/ 100 5-point scale 3.5 / 5
This review was completed under FSRS 1.0. Dimension scores are not converted; it is queued for re-review under FSRS 2.0.
1 2 3 4 5 6
1Utility11 / 20 · 2.8/5
2Reliability14 / 20 · 3.5/5
3Safety24 / 25 · 4.8/5

Explicitly requires drafts for human review only, never auto-posting to GitHub, and requires verifying claims against the code (citing path:line) rather than fabricating them when unverifiable — a well-designed safety constraint for the high-risk scenario of speaking as the maintainer.

4Evidence6 / 15 · 2.0/5
5Usability7 / 10 · 3.5/5
6Maintenance7 / 10 · 3.5/5
Evidence confidence:Low Reviewed Jul 17, 2026
Review evidence [1]
See the full review method →

What does this skill do, and when should you use it?

This is a Claude Skill inside the agent-service-toolkit repository that drafts GitHub issue and PR replies in the authentic voice of maintainer JoshuaC215. Before writing anything, it reads the full issue/PR context — body, comments, code, CI status — and maps cross-referenced items into clusters so replies stay consistent with each other and with the maintainer's prior statements. Output is always a draft plus rationale for human review; it never posts, merges, closes, or labels on its own. It also encodes the maintainer's actual tone, stock phrases, accept/decline heuristics for contributions, and a documented gh CLI workflow for posting with clean attribution.

Reads full context for each issue/PR (body, all comments, code/diff, CI status, fork status, whether the maintainer already replied); pulls cross-references in both directions to build a cluster map of issues, linked PRs, and sibling PRs, and reads the maintainer's prior comments across the whole cluster; classifies each cluster and recommends accept/decline/defer; writes reply drafts in Joshua's voice with verified file:line citations or runnable snippets; produces a structured batch report with hyperlinked items, request summaries, research assessments, decisions needed, and paste-ready drafts; after explicit human approval, posts via gh CLI using --body-file and -R JoshuaC215/agent-service-toolkit, verifying the gh login identity is JoshuaC215 first.

  1. A solo maintainer using an AI assistant for biweekly batch triage of issues and PRs who needs a batch of consistently voiced drafts
  2. Replying to months-old piled-up issues with apologetic, well-mannered drafts
  3. Reviewing bulk AI-generated PRs and verifying whether the claimed bug is real, reproducible, and free of breaking defaults
  4. Multiple contributors attacking the same feature — clustering first so positions don't contradict each other
  5. Politely closing stale issues or declining features while inviting upvotes to gauge real demand

What are this skill's strengths and limitations?

Pros
  • Clear draft-only safety boundary — never mutates GitHub state without explicit approval
  • Cluster-based triage: reads the whole feature cluster and the maintainer's prior stances before drafting, preventing contradictions
  • Concrete voice guide, stock phrases, and real accept/decline heuristics make drafts genuinely human-sounding
  • Detailed local gh CLI posting workflow ensures clean attribution (no AI badge or footer)
  • Explicit handling of tricky cases: fork PRs, unreproducible bugs, AI-generated PRs
Limitations
  • Tightly bound to one repo (JoshuaC215/agent-service-toolkit) and one maintainer's personal voice; porting elsewhere requires substantial rewriting
  • References workflow files like DRAFT_RESPONSES_REPORT.md and docs/maintenance/ that may not exist outside this repo
  • Requires gh CLI authenticated as the maintainer — a high bar for credentials and permissions
  • Its current (2026) stance is deliberately conservative, defaulting toward declining or deferring contributions — adopters should know this

How do you install this skill?

The skill lives at .claude/skills/maintainer-response/ in the repository (obtained by cloning). Run git clone https://github.com/JoshuaC215/agent-service-toolkit.git, then copy the .claude/skills/maintainer-response/ folder into your Claude Code skills directory. The README covers the collection's setup; finer standalone-install steps for this one skill are not documented in the source.

How do you use this skill?

Run in Claude Code as the maintainer, e.g. "respond to issue 290", "review the open PRs", "draft a reply to this bug report", or a batch triage prompt. The skill researches and clusters first, then reports per item: summary, assessment, decision needed, and a quoted draft. Only after you explicitly clear a named item does it post via gh CLI (with --body-file and always -R JoshuaC215/agent-service-toolkit), after confirming gh is authenticated as JoshuaC215.

FAQ

Does it post comments to GitHub on its own?
No. It only produces drafts and a report. Posting happens only after the human explicitly clears a named item, via gh CLI authenticated as the maintainer, with an identity check first.
Can I use it on my own repository?
Not as-is. The tone guide, lint commands, decision heuristics, and gh commands are hardcoded for agent-service-toolkit and Joshua's personal style; adapting it requires rewriting.
What permissions and dependencies does it need?
Network access to GitHub, local filesystem access, and gh CLI authenticated as JoshuaC215 for posting. Drafting alone requires only read access to the repository.
How does it treat AI-generated PRs?
With skepticism: it verifies the claimed bug is real and reproducible, checks for breaking defaults and real tests with passing CI, and politely asks the author to confirm the actual repro.

More skills from this repository

All from JoshuaC215/agent-service-toolkit

Related skills