transcript-inspector

A specialist for investigating why a coding assistant’s stop hook failed. A stop hook is an automatic check that can prevent the assistant from ending before required work is explained or completed.

In plain words
What is it for?
Use it when a stop hook did not block an early stop, when manual “continue” was needed, or when a transcript reports the wrong continuation result.
Why use it?
It helps identify why the assistant stopped unexpectedly by examining session transcripts and daemon logs.

Agent for Claude Code

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/edmonds-commerce-limited/claude-code-hooks-daemon/transcript-inspector
Clone the repo
git clone --depth 1 https://github.com/Edmonds-Commerce-Limited/claude-code-hooks-daemon

Made for: Claude Code.

Per session 49 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,087 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 $0.00049 $0.03087
Opus 5 $0.00024 $0.01543
Sonnet 5 $0.00010 $0.00617
Haiku 4.5 $0.00005 $0.00309

Measured 2d ago against content hash 06d77bc0d486, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

transcript-inspector 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.

.claude/agents/transcript-inspector.md · 347 lines

How it starts

The opening of the file, as written. The whole thing — 347 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Transcript Inspector - Stop Hook Dogfooding Diagnostics

Purpose

Diagnose stop hook failures by reading Claude Code session transcripts and cross-referencing with daemon logs. Primary use case: "Claude stopped when it should have been blocked" or "the stop hook isn't working."

How to Use This Agent

Invoke when:

  • Claude stops without a STOPPING BECAUSE: prefix and the daemon did not block it
  • A user had to type "go" or "continue" manually (symptom of broken stop hook)
  • preventedContinuation: false appears in transcript when it should be true
  • Any stop hook dogfooding regression

Step 1: Find the Current Transcript

# List all transcript files for this project, sorted by modification time (newest last)
ls -lt /root/.claude/projects/-workspace/*.jsonl 2>/dev/null | head -5

# The most recent file is the current session
TRANSCRIPT=$(ls -t /root/.claude/projects/-workspace/*.jsonl 2>/dev/null | head -1)
echo "Transcript: $TRANSCRIPT"

The project path encoding: /workspace-workspace. If running from a different path, adjust accordingly.

Step 2: Locate stop_hook_summary Entries

These are the definitive outcome records for every stop hook invocation.

#!/usr/bin/env python3
import json
import sys

transcript = sys.argv[1]  # pass transcript path as arg

with open(transcript) as f:
    lines = f.readlines()

print(f"Total lines: {len(lines)}\n")

for i, line in enumerate(lines):
    try:
        e = json.loads(line)
    except json.JSONDecodeError:
        continue

    # Stop hook outcome record
    if e.get('subtype') == 'stop_hook_summary':
        print(f"Line {i}: stop_hook_summary")
        print(f"  preventedContinuation : {e.get('preventedContinuation')}")
        print(f"  level                 : {e.get('level')}")
        print(f"  hasOutput             : {e.get('hasOutput')}")
        print(f"  hookErrors            : {e.get('hookErrors')}")
        print(f"  raw                   : {json.dumps(e, indent=2)}")
        print()

    # Stop hook progress event (hook firing)
    if (e.get('type') == 'progress'
            and isinstance(e.get('data'), dict)
            and e['data'].get('type') == 'hook_progress'
            and e['data'].get('hookEvent') == 'Stop'):
        print(f"Line {i}: Stop hook FIRING (progress event)")
        print(f"  hookName : {e['data'].get('hookName')}")
        print(f"  toolUseID: {e.get('toolUseID')}")
        print()

Read the full file on GitHub · 347 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 · 347 lines · 49 tokens per session scan A 06d77bc0d486

Subscribe to this mod's changes

transcript-inspector is an agent published in the GitHub repository Edmonds-Commerce-Limited/claude-code-hooks-daemon (3 stars, last pushed 2d ago), licensed MIT. It adds 49 tokens to every session and 3,087 once invoked, about $0.0002 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 2026-08-31.

Related

Other agents, from other repositories

api-tester

Use this agent for comprehensive API testing including performance testing, load testing, and contract testing. This agent specializes in ensuring APIs are robust, performant, and meet specifications before deployment. Examples:\n\n \nContext: Testing API performance under load.

ccplugins/awesome-claude-code-plugins · 54 tokens

backend-architect

Use this agent when designing APIs, building server-side logic, implementing databases, or architecting scalable backend systems. This agent specializes in creating robust, secure, and performant backend services. Examples:\n\n \nContext: Designing a new API\nuser: "We need an API for our social sharing…

ccplugins/awesome-claude-code-plugins · 287 tokens

claim-verifier

Adversarially verifies an ARBITRARY free-text claim before it is trusted. Runs an in-harness panel of N independent skeptics, each told to REFUTE the claim (default-refute-if-uncertain, fail-closed), then returns a STRUCTURED majority-vote verdict (HOLDS / REFUTED, how many refuted, the threshold, per-skeptic…

dwarvesf/dwarves-kit · 168 tokens

meta-agent

The agent that drafts agents. From a one-line description, drafts a new subagent definition OR a new mega-goal sub-goal file, matching the kit's exact frontmatter + structure. Output is always a DRAFT for human review; it never self-installs or self-runs.

dwarvesf/dwarves-kit · 62 tokens

task-verifier

Verifies a completed task against its spec acceptance criteria. Run after each worker subagent completes a task. Read-only -- cannot modify the codebase.

dwarvesf/dwarves-kit · 34 tokens

brief-reviewer

Statically reviews a design brief or requirement (DECISION-BRIEF.md, a spec's Problem/Context section, or an equivalent requirement doc) for clarity, completeness, and testability before it hardens into a spec. Read-only -- cannot edit the brief.

dwarvesf/dwarves-kit · 58 tokens