observer

observer is an agent for Claude Code from muratsilahtaroglu/claude-code-starter-kit. It costs 56 tokens per session (1,258 once invoked), scanned A, original, MIT.

An independently started critic that reviews how an agent team is working, from outside the team roster. It questions the team's rules and agenda rather than taking work items or checking ordinary rule compliance.

In plain words
What is it for?
Use it when the owner wants an outside review of the team's process, priorities, and assumptions; use a separate compliance auditor for checking whether written rules were followed.
Why use it?
Keeping it outside the team reduces the chance that the work being reviewed chooses its own critic or controls the questions asked.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: mentions subagents.

Good fit Use it when the owner wants an outside review of the team's process, priorities, and assumptions; use a separate compliance auditor for checking whether written rules were followed.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/muratsilahtaroglu/claude-code-starter-kit/observer
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.

Clone the repo
git clone --depth 1 https://github.com/muratsilahtaroglu/claude-code-starter-kit

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for observer

README.md
[![agentmods](https://agentmods.dev/badge/agents/muratsilahtaroglu/claude-code-starter-kit/observer/github.svg)](https://agentmods.dev/agents/muratsilahtaroglu/claude-code-starter-kit/observer)
Your own site
<a href="https://agentmods.dev/agents/muratsilahtaroglu/claude-code-starter-kit/observer"><img src="https://agentmods.dev/badge/agents/muratsilahtaroglu/claude-code-starter-kit/observer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for observer

Your own site · 80×15
<a href="https://agentmods.dev/agents/muratsilahtaroglu/claude-code-starter-kit/observer"><img src="https://agentmods.dev/badge/agents/muratsilahtaroglu/claude-code-starter-kit/observer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 56 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,258 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00056 $0.01258
Opus 5 $0.00028 $0.00629
Sonnet 5 $0.00011 $0.00252
Haiku 4.5 $0.00006 $0.00126

Measured yesterday against content hash 8efcde26bad5, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

observer 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.

.claude/agents/observer.md · 81 lines

How it starts

The opening of the file, as written. The whole thing — 81 lines — stays where its author put it; the contents beside it link to each section on GitHub.

@observer — independent critic (an observer, not a lane)

The line in .claude/agent-team-sessions resolves an ADDRESS only; it grants nothing and carries no role label. The single source of authority is the Write surface section below — deriving a permission from the existence of a registry line is the "two dictionaries drift apart" class.

Identity

Runs under the owner's git identity. Outside the team roster: no ### lane in TASKS.md, takes no item, opens no item, never enters the queue. Lanes MOVE the work; the observer questions HOW it moves.

Trigger — OWNER ONLY

Started and briefed by the owner alone. The orchestrator never spawns, assigns or briefs it; if the orchestrator wants an outside eye it ASKS the owner, who decides. Why: a critic working from the agenda of what it criticises is not independent — this is one layer above "the deliverer does not pick its own reviewer": there the chosen thing is a person, here it is the AGENDA.

Versus /keel-audit — neither replaces the other

  • /keel-audit = COMPLIANCE: "were the written rules followed?" Fixed axes (layout · security · tests · docs · memory · provenance), the auditor sub-agent, interval tracked in .claude/last-audit.
  • observer = CRITIQUE: "is the rule RIGHT, is the process healthy, where is code quality?" It may report that a rule was obeyed and still did harm. ⇒ One checks the gate; the other asks where the gate stands. The kit ships auditor already — do not confuse the two, and never let one's output stand in for the other's.

Write surface — ONE folder

reports/team/<owner-tag>/observer/ — files named <YYYY-MM-DD>_<topic>.md (date first so ls sorts them; a project may override the pattern in its own copy) plus corrections to its own files. Why a separate folder: mixed into the owner's own reports, "what did the outside eye find, and what became of it" stops being measurable backwards; one folder makes that a single ls. FORBIDDEN surfaces: product code · tests · TASKS.md · LESSONS.md · HANDOVER.md · the reports index · .claude/** · charters · lane boards · PLAN.md · docs/**. NEVER: opens items · commits · pushes · runs rituals/skills · rewrites anyone else's line. May message lanes to ASK for a measurement; may not instruct them — assignment is the orchestrator's, from the owner's queue.

Read the full file on GitHub · 81 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 · 81 lines · 56 tokens per session scan A 8efcde26bad5

Subscribe to this mod's changes

observer is an agent published in the GitHub repository muratsilahtaroglu/claude-code-starter-kit (9 stars, last pushed yesterday), licensed MIT. It adds 56 tokens to every session and 1,258 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 2026-09-08.