AGENT SKILL

resume

Resume an interrupted research-agent run from .research/. Re-launches the researcher to reload its structured state and continue from the current focus, optionally on a named branch or with a new budget. Use after a session restart, crash, or explicit pause.

License 许可
MIT
最近更新
2026-10-23
Stars
★ 0
Forks
0

01 前言 Front Matter

name
resume
description
Resume an interrupted research-agent run from .research/. Re-launches the researcher to reload its structured state and continue from the current focus, optionally on a named branch or with a new budget. Use after a session restart, crash, or explicit pause.
argument-hint
[branch-name] [budget: N cycles, H hours]

03 SKILL.md 正文

Resume Research

Resume the research run per: $ARGUMENTS

Step 1: Resolve the Run Root

  • If the user names a branch, the run root is .research/branches/<name>/; it must exist (Glob tool).
  • Otherwise the run root is .research/.
  • .research/goal.md marks a research run. If it is absent, there is nothing to resume: suggest /research-agent:research.

Parse an optional new budget (cycle cap and/or hours). Without one, the agent keeps the budget persisted in the run root's run.json.

Step 2: Refuse if a Researcher Is Live

Read <run root>/run.json if present. The run is live when status is running or stopping and heartbeat is less than stale_after_minutes old (compare against date -u +%Y-%m-%dT%H:%M:%SZ). If live, do not launch a second researcher: tell the user one is already working (with the heartbeat age) and point at /research-agent:status or /research-agent:synthesize. The single exception: the user states the earlier session is gone (for example, it was closed moments ago) and confirms they want to proceed (AskUserQuestion).

Step 3: Handle a Concluded Run

If <run root>/synthesis.md exists, the run has concluded. Ask the user (AskUserQuestion) whether to reopen it. On yes, include <reopen/> in the launch prompt (the agent archives the old synthesis); suggest a larger budget if the previous one ran out. On no, point at synthesis.md and stop.

Step 4: Preview

Show a one-paragraph preview of what the researcher will pick up:

Resuming research on "{goal}". Current focus: {current_focus}. Last cycle was {N} on {date}: {modality} -> {result}. Open sub-problems: {count}. Untested hypotheses: {count}. Budget: {used}/{max} cycles.

Sources, structured first:

  • goal: the body of .research/goal.md (below its frontmatter)
  • current focus and counts: the YAML frontmatter of <run root>/state.md
  • last cycle: the last line of <run root>/cycles.jsonl
  • budget: <run root>/run.json

Legacy fallback (a run from an older plugin version: state.md without frontmatter, or no cycles.jsonl): use the prose "Current focus" section of state.md and the last ### Cycle entry of log.md, and note that the researcher will migrate the run to the structured format. If state.md is missing entirely, say the researcher will rebuild it from goal.md and log.md.

Step 5: Launch the Researcher

Spawn the researcher (Agent tool, subagent_type: research-agent:researcher) with only these tags; the agent's resume-mode initialization defines the steps:

<mode>resume</mode>
<branch>{name}</branch>                       (only when resuming a branch)
<budget max_cycles="{N}" max_hours="{H}"/>    (only when the user gave a new budget)
<reopen/>                                     (only when the user confirmed a reopen)
<working_directory>
{absolute path to current working directory}
</working_directory>

Do not pass isolation: a resumed branch reuses the worktree recorded in its run.json.

Step 6: Inform the User

Researcher resumed from {run root}. Use /research-agent:status to check progress, or /research-agent:synthesize to ask it to conclude.

04 同仓库更多 Skills

Abdulqudduz
Abdulqudduz/claude-anvil

worldsmith-methodology

Use when working in a project with a worldbuilding documentation ecosystem (docs/ or lore/ directory). Provides the documentation-first methodology: canonical hierarchy, doc roles (timeline, lore, characters, systems), propagation discipline, and the dual canonical and exploratory workflow.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

prose-craft

Use when writing fiction, editing fiction prose, reviewing chapters, or drafting scenes. Covers prose craft, dialogue mechanics, scene structure, and common failure modes in AI-assisted fiction. Trigger phrases: "write fiction", "edit chapter", "prose review", "scene draft", "dialogue check", "fiction craft", "edit this scene", "review my prose".

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

vista

This skill should be used when the user asks about research directions, open problems, future work, or follow-up research from the academic literature. Trigger phrases include "open problems in X", "what's next for Y", "future work for paper Z", "research directions on T", "salient follow-up research", "what should I work on next in", "find me follow-up research unrelated to my prior work", "broad survey of W", "neglected directions in V". Routes the request to the right Vista MCP tool, reads back the structured paper sections, and synthesizes research directions in the conversation.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

synthesize

Make a research-agent run conclude and write synthesis.md (outcome, key findings, failed approaches, open questions, recommendations). Signals a live researcher to wrap up, or launches one in synthesis mode when none is running. Use when results are good enough or the agent is stalling.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

status

Show the current state of a research-agent run from .research/: liveness, cycles and budget, sub-problem and hypothesis statuses, eval trend, current focus, and branches. Read-only.

★ 0
Abdulqudduz
Abdulqudduz/claude-anvil

research

Launch an autonomous research agent that iterates toward a goal through proofs, code, simulations, tests, and counterexample search. Fire-and-forget: the agent runs on its own and writes all results to .research/ in the target project.

★ 0