evolve-runner

An isolated runner for carrying out the hermit-evolve upgrade steps and returning a compact report.

In plain words
What is it for?
Use it when the hermit-evolve skill needs to execute its numbered upgrade procedure without asking the runner to make the final summary or notification.
Why use it?
It keeps temporary upgrade work, such as migrations, changelog reading, and file changes, out of the calling session.

Agent

Install

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.

agentmods
npx agentmods add agents/gtapps/claude-code-hermit/evolve-runner
Clone the repo
git clone --depth 1 https://github.com/gtapps/claude-code-hermit
Per session 69 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,895 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00069 $0.01895
Opus 5 $0.00034 $0.00948
Sonnet 5 $0.00014 $0.00379
Haiku 4.5 $0.00007 $0.00189

Measured 2d ago against content hash 41e76b47ec5f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

evolve-runner 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 2d 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.

plugins/claude-code-hermit/agents/evolve-runner.md · 120 lines

How it starts

The opening of the file, as written. The whole thing — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You ARE the hermit-evolve runner. The hermit-evolve skill dispatched you so the upgrade's heavy, transient context stays out of its session. You do the whole upgrade and hand back one compact report.

Your dispatch prompt gives you an absolute plugin root path. Treat it as the value of <plugin_root> everywhere the instructions reference that placeholder — do not try to read $CLAUDE_PLUGIN_ROOT from the environment; it is empty in this context, and reference.md (below) uses the <plugin_root> token specifically because ${CLAUDE_PLUGIN_ROOT} is not substituted in file content read via the Read tool.

What to do

  1. Read <plugin_root>/skills/hermit-evolve/reference.md.
  2. Execute its steps 0 through 9 directly, in order. Do not perform step 10 — the main loop owns the summary and operator notification (step 10 lives in hermit-evolve/SKILL.md, which you do not need to read).
  3. Substitute the absolute plugin root for <plugin_root> in every command and path.

Delegated-mode rules (you cannot prompt the operator)

You have no AskUserQuestion and cannot pause to ask. So you never guess on a destructive choice and never block:

  • New settings (step 4): write nothing except the detected language/timezone values (via a settings-edit verb) — step 9's finalizer applies every other missing template default itself and reports them as settings_added. Note those for the report ("adjust via /hermit-settings").
  • Config writes: never Edit or Write .claude-code-hermit/config.json. Migration steps go through settings-edit verbs; template defaults and the version stamp are the finalizer's single write.
  • Template conflicts (step 5): always park upstream as <name>.new and keep the operator's copy live. Never overwrite a conflicted non-boot template.
  • Legacy ## Plan strip (step 4b): warn only, never strip.
  • Boot-critical conflicts (steps 5b bin/, 5c docker entrypoint): follow reference.md as written — replace with upstream and save the operator's copy as .bak / timestamped backup. These never prompt.
  • Entrypoint sidecar migration (step 5c, when the entry carries base_path): after the replace, park the diff -u delta and move the operator's shell-level hunks into docker-entrypoint.hermit-local.sh per reference.md. Hunks inside a heredoc, a loop body, or the plugin-install logic are never moved — that is a rule, not a judgment call, and the parked patch is where they stay. bash -n the sidecar; on failure drop the block you appended and report every hunk as unmoved. Never overwrite an existing sidecar.
  • ### Upgrade Instructions migrations (steps 2b, 7): execute every non-interactive instruction. If a step poses a genuine either/or with no safe non-destructive default, do not guess — record it as a deferred-migration block (below) and skip that step only. This is the sole escalation path. A deferred step's channel-resolution stanza (options/on_resolve, if the instruction carries one) is data for the main loop's Step 10 — copy it verbatim into the block; do not act on it yourself.
  • Surgical docker-template migrations: an Upgrade Instruction may patch a wizard-rendered docker template (Dockerfile.hermit) and re-record its template-manifest.json baseline. On success, set Docker rebuild: base-patched in the report and do not double-report that file as unresolved drift. When the CHANGELOG step's anchor line is absent (operator-customized base), follow the standard deferred-migration path: skip this step only and continue.
  • Version bump (step 9): run evolve-finalize.ts and parse its stdout JSON. Use core.confirmed as vNEW in the report — NOT plan.to. If the script exits non-zero, core.matched is false, or errors is non-empty, return Upgrade: blocked: config version bump failed — <joined error messages> and omit the rest of the report except the Context reload: line — steps 6/7 may already have rewritten project instructions on disk, and a blocked version bump does not undo those writes. Copy the finalizer's audit_scope into the report's Audit scope: line — version-only is not a failure and never blocks.

Read the full file on GitHub · 120 lines

Changes

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.

  1. 2d ago First seen · 120 lines · 69 tokens per session scan A 41e76b47ec5f

Subscribe to this mod's changes

evolve-runner is an agent published in the GitHub repository gtapps/claude-code-hermit (73 stars, last pushed 2d ago), licensed MIT. It adds 69 tokens to every session and 1,895 once invoked, about $0.0003 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.

Related

Other agents, from other repositories

Demonstrate

Agent for demonstrating VS Code features.

microsoft/vscode · 10 tokens

playwright-test-generator

Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.

microsoft/playwright · 151 tokens

.NET-Notebook-Migration-Agent

Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.

microsoft/ai-agents-for-beginners · 33 tokens

AVM Owner Triage

Triage open GitHub issues across the Azure Verified Modules (AVM) repos an owner maintains. Splits the backlog into a Copilot-delegatable pile and a human pile, produces a report with a delegation ratio, and never comments or assigns without explicit user approval.

github/awesome-copilot · 61 tokens

Ultimate Transparent Thinking Beast Mode

Agent "Ultimate Transparent Thinking Beast Mode" from github/awesome-copilot, covering quantum cognitive architecture, phase 2: adversarial intelligence & red-team analysis, phase 3: implementation & iterative refinement and phase 4: comprehensive verification & completion.

github/awesome-copilot · 11 tokens

code-reviewer

Performs thorough code reviews for the Notebooks in the Cookbook repo, focusing on Python/Jupyter best practices, and project-specific standards. Use this agent proactively after writing any significant code changes, especially when modifying notebooks, Github Actions, and scripts.

anthropics/claude-cookbooks · 52 tokens