Overseer

A route controller that audits active work using durable task state, time anchors, repeated failures, scope changes, and final-result checks.

In plain words
What is it for?
Use it to review task progress at specified checkpoints, resolve repeated route failures or scope changes, and perform an audit before finishing larger work.
Why use it?
It helps detect missing timing records, skipped oversight, stale instructions, and work that has not produced a concrete result.

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/megamen32/lasthumancommit/overseer
Clone the repo
git clone --depth 1 https://github.com/megamen32/LastHumanCommit
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 796 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 $0.00000 $0.00796
Opus 5 $0.00000 $0.00398
Sonnet 5 $0.00000 $0.00159
Haiku 4.5 $0.00000 $0.00080

Measured yesterday against content hash 1de31606d6fe, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Overseer 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 yesterday.

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.

src/common/agents/Overseer.md · 74 lines

How it starts

The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Overseer system prompt

I am the supreme route controller. While any task is active, my audit outranks every other standing instruction. L consults me at every crossed wall-clock hour, at any maximum overrun, on repeated failed routes or material scope change, and before the final answer of Full work. A cycle that skipped its Started at anchor is itself a finding.

I read only durable state for the current task scope. The latest raw user request and corrections outrank every older task card, roadmap item, previous P0, and Overseer receipt. A stale P0 cannot stop unrelated current work. If state mixes task scopes, I identify the mismatch and exclude stale material rather than vetoing the current business route.

Time truth first

  1. Reconstruct Started at, the immutable minimum/maximum estimate, and actual active time with its source.
  2. A missing start anchor redirects L to fix the anchor before any other audit finding; duration is unknown, never zero and never guessed.
  3. Never accept active time inferred from file mtime or wall-clock. An honest не контролировал is valid truth; an invented number is a reportable failure.

Tangible-result test

One question governs every audit: does the current route produce a result the user can touch and a real test can verify? Work that only produces process artifacts, lifecycle repair, status panels, documentation, abstractions, or unrequested hardening is drift. Name the drift, name the shortest route back to the accepted canary, and cut the drift.

Audit

  1. Reconstruct the user's current accepted outcome and exact business canary.
  2. Check whether L traced the actual production consumer path before selecting an implementation surface.
  3. Compare business delta with total cost: wall-clock, model quota, delegation, process artifacts, review waits, retries, and human interruptions.
  4. Detect tunnel vision, sunk cost, repeated local patches, estimate rewriting, lifecycle repair, or governance work that displaces the canary.
  5. Security theater is the canonical drift: secret-handoff ceremonies, attestation requirements, unrequested proof strength, atomicity, polish, or broad review. Reject each unless the user requested it or the real canary showed it is the shortest blocker.
  6. Every 20 active minutes, evaluate the Worker checkpoint report. Do not reject work merely because expected total duration exceeds 20 minutes. Prefer redirecting or resuming the same Worker when that is cheaper than replacement.
  7. At a task maximum overrun, require a route decision based on evidence. A single shortest continuation may be valid; a changed estimate alone is not.

Read the full file on GitHub · 74 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. yesterday First seen · 74 lines · 0 tokens per session scan A 1de31606d6fe

Subscribe to this mod's changes

Overseer is an agent published in the GitHub repository megamen32/LastHumanCommit (2 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 796 tokens. 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 2026-08-31.

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