ワンクリックで
herdr-agent-observation
Observe Herdr peers safely with native read-only herdr commands.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Observe Herdr peers safely with native read-only herdr commands.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Authenticated browser QA on the Voyager dev server — mint real AuthKit sessions for seeded test personas via /api/dev/login, then drive the UI with Playwright. Use whenever QA, visual review, or E2E verification needs a logged-in user.
Orchestrate parallel work across worker agents in herdr panes (omp by default; fable/claude workers on explicit user request) — spawn or discover workers, dispatch assignment packets, monitor liveness, and verify results before committing. Use whenever a code change should be delegated instead of implemented firsthand (delegation-first policy), e.g. /herdr-orchestrate 3 to spawn three workers, or /herdr-orchestrate p17,p18,p19 to use existing panes.
Initialize or repair AOC context, memory, and tasks safely.
Ponytail review, audit, and debt workflows for over-engineering review and shortcut ledger scans. Use for /ponytail review, /ponytail audit, /ponytail debt, or when asked what to delete, simplify, or track as ponytail debt.
Anime.js v4 API correctness, imports, targeting, value mapping, refresh/revert semantics, and sane implementation defaults. Use when writing or fixing Anime.js code.
Audit Anime.js work for performance, reduced-motion behavior, engine settings, scroll readability, SVG costs, and cleanup discipline.
| name | herdr-agent-observation |
| description | Observe Herdr peers safely with native read-only herdr commands. |
Use native herdr commands for observation. Default to read-only commands:
herdr agent list
herdr agent get <target>
herdr agent read <target> --source recent --lines 80
herdr agent explain <target>
herdr agent wait <target> --status idle --timeout 60000
herdr workspace list
herdr workspace get <workspace_id>
herdr tab list [--workspace <workspace_id>]
herdr tab get <tab_id>
herdr pane list [--workspace <workspace_id>]
herdr pane get <pane_id>
herdr pane read <pane_id> --source recent --lines 120
Prefer agent commands when Herdr has identified an OMP/agent label. Use pane commands for raw terminal observation, unknown processes, or when you already have a pane id.
If Herdr reports an agent session path or session id, use herdr agent explain <target> first. Read OMP session files directly only when native herdr agent read, herdr pane read, and herdr agent explain do not provide the needed transcript context.
When reading OMP session files:
Do not use control or input commands unless the operator explicitly asks for that exact action and target:
herdr agent send <target> <text>
herdr agent focus <target>
herdr agent start <name> -- <argv...>
herdr pane send-text <pane_id> <text>
herdr pane send-keys <pane_id> <key> [...]
herdr pane run <pane_id> <command>
herdr pane focus --direction ...
herdr tab focus <tab_id>
herdr workspace focus <workspace_id>
Never start, focus, type into, or send keys to a peer just to observe it. Observation is list/get/read/explain/wait plus pane/tab/workspace listing and reads.