observe-agent-runs

observe-agent-runs is a skill for Claude Code from EliBarak12/Elliot. It costs 49 tokens per session (887 once invoked), scanned B, original, MIT.

A setup that sends a coding-agent session to the Elliot Agent Console while it runs. The session includes prompts, model reasoning, tool calls, final answers, and MCP tool calls.

In plain words
What is it for?
Use it to install or check a trace hook for Claude Code, Codex, or Cursor. It posts the session to an Elliot runtime, normally at localhost port 3001, or to a configured remote runtime URL.
Why use it?
Without the trace hook, the console sees only tool calls and not the surrounding conversation or reasoning. Streaming the full session makes the agent’s work easier to inspect.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: reads .claude/ paths; mentions Claude Code; mentions Codex.

Part of the elliot plugin — 10 skills, 2 hooks, 1 MCP server shipped together

Good fit Use it to install or check a trace hook for Claude Code, Codex, or Cursor. It posts the session to an Elliot runtime, normally at localhost port 3001, or to a configured remote runtime URL.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/elibarak12/elliot/observe-agent-runs
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.

Any agent
npx skills add EliBarak12/Elliot --skill observe-agent-runs
Clone the repo
git clone --depth 1 https://github.com/EliBarak12/Elliot

Made for: Claude Code.

Or install elliot, the plugin that ships this one along with the rest of its 10 skills, 2 hooks, 1 MCP server.

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 observe-agent-runs

README.md
[![agentmods](https://agentmods.dev/badge/skills/elibarak12/elliot/observe-agent-runs/github.svg)](https://agentmods.dev/skills/elibarak12/elliot/observe-agent-runs)
Your own site
<a href="https://agentmods.dev/skills/elibarak12/elliot/observe-agent-runs"><img src="https://agentmods.dev/badge/skills/elibarak12/elliot/observe-agent-runs/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 observe-agent-runs

Your own site · 80×15
<a href="https://agentmods.dev/skills/elibarak12/elliot/observe-agent-runs"><img src="https://agentmods.dev/badge/skills/elibarak12/elliot/observe-agent-runs.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 49 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 887 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. 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 $0.00049 $0.00887
Opus 5 $0.00024 $0.00443
Sonnet 5 $0.00010 $0.00177
Haiku 4.5 $0.00005 $0.00089

Measured 11d ago against content hash 9fa002dfd6e6, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade B, and why

observe-agent-runs scanned grade B with 1 finding 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 11d 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.

Reads agent configuration directoriesmediumAgent snooping

.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.

`~/.claude/settings.json` for Claude Code), returns the config path it
skills/observe-agent-runs/SKILL.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.

Observe Agent Runs

You are wiring the Elliot trace hook into a coding agent's config so the full session — prompt, model reasoning, tool calls, and final answer — appears in the Agent Console. Without the hook the Console only sees MCP tool calls, which is half the story: the why lives in the prompt and the reasoning.

This is principle #4 — every agent session is observable — operationalised.

Prerequisite

Elliot must be running (make dev). The hook posts to the connector-runtime at http://localhost:3001/v1/trace/ingest, so the runtime port (3001) needs to be reachable from the agent's machine. If the user runs Elliot on a remote host, set ELLIOT_RUNTIME_URL in their shell before installing.

Steps

1. Check current state

Call elliot_trace_hook_status. It returns the runtime URL the hook will post to and, for each supported harness (claude-code, codex, cursor), whether it's installed today and the config path it lives in.

If the harness the user uses already shows installed: true, jump to step 3 and tell them what they'd see when they next run that harness.

2. Install for the right harness

Ask the user which agent they want streamed into Elliot (or detect from context — if they're talking to you in Claude Code, default to claude-code).

Call elliot_install_trace_hook(harness="<harness>") with one of: claude-code, codex, cursor.

It writes a hook entry into the harness config (e.g. ~/.claude/settings.json for Claude Code), returns the config path it touched, and the runtime URL the hook will post to. Tell the user:

"Installed the trace hook into <config_path>. Restart <harness>; its next prompts, reasoning, tool calls, and final answers will stream into the Agent Console at http://localhost:5173/agents."

3. Verify it works

After the user restarts the harness and makes one call, ask them to refresh the Agent Console. They should see a new run entry that includes the prompt text and the model's reasoning, not just the tool calls.

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. 11d ago First seen · 81 lines · 49 tokens per session scan B 9fa002dfd6e6

Subscribe to this mod's changes

observe-agent-runs is a skill published in the GitHub repository EliBarak12/Elliot (11 stars, last pushed 10d ago), licensed MIT. It adds 49 tokens to every session and 887 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.