bob-debug

A read-only debugger for completed or stuck Hacker Bob security-research sessions. It reviews telemetry, saved artifacts, and selected transcript sections to assess pipeline quality, drift, failures, and possible improvements.

In plain words
What is it for?
Use it to inspect the latest or a named session, perform a deeper review, or compare two session attempts.
Why use it?
It helps find what went wrong without changing the session, contacting the target, or producing a formal evaluation. It can also compare two attempts.

Skill for Claude CodeCodex

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 skills/vmihalis/hacker-bob/bob-debug
Any agent
npx skills add vmihalis/hacker-bob --skill bob-debug
Clone the repo
git clone --depth 1 https://github.com/vmihalis/hacker-bob

Made for: Claude Code, Codex.

Per session 21 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,842 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 $0.00021 $0.01842
Opus 5 $0.00010 $0.00921
Sonnet 5 $0.00004 $0.00368
Haiku 4.5 $0.00002 $0.00184

Measured 3d ago against content hash 945f2ea98ff5, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

bob-debug 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 3d 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/skills/bob-debug/SKILL.md · 112 lines

How it starts

The opening of the file, as written. The whole thing — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the read-only post-session debugger for Bob. Review a completed or stuck Hacker Bob session and explain pipeline quality, drift, failures, and concrete improvements. Do not evaluate, verify, grade, report, mutate state, or interact with the target.

Input: $ARGUMENTS (--last, no args, <target_domain>, optionally plus --deep, or --diff-attempts <prev> <curr> for cross-attempt v2 inspection)

Hard Rules

  • Read-only only. Never call mutating MCP tools, never write files, never merge waves, never transition phases, never update auth, never write reports, and never use HTTP scan or browser/target interaction tools.
  • Do not use the Task tool by default. Debug locally from telemetry, MCP reads, artifacts, and narrow transcript windows.
  • Do not create a debug bundle in v1. Print the assessment only.
  • Telemetry MCPs are the first source of truth. Artifacts and transcripts are supporting evidence.

Argument Handling

  • No args or --last: inspect the latest local session under ~/hacker-bob-sessions.
  • <target_domain>: inspect that specific session directory.
  • --deep: additionally inspect Claude transcript windows around flagged issues.
  • --diff-attempts <prev> <curr>: cross-attempt v2 verification diff. Each token is either an archive id from bob_read_verification_context.data.archived_attempts[*].attempt_id, or the literal string current for the live attempt. Calls bob_diff_verification_attempts({ target_domain, attempt_a: <prev>, attempt_b: <curr> }) and prints the snapshot / adjudication / final hash matches plus the per-file divergence (only-in-a, only-in-b, and content-changed entries with truncated 16-char hashes). Use this to explain why a re-verification produced different results across attempts.
  • If both a domain and --deep are present, debug that domain deeply. If multiple non-flag tokens are present, stop and ask for one target domain.

Latest-session detection must pick the newest target directory by pipeline-events.jsonl mtime. If no pipeline event file exists, fall back in order to state.json, grade.json, report.md, then directory mtime.

Read the full file on GitHub · 112 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. 3d ago First seen · 112 lines · 21 tokens per session scan A 945f2ea98ff5

Subscribe to this mod's changes

bob-debug is a skill published in the GitHub repository vmihalis/hacker-bob (97 stars, last pushed 4d ago), licensed Apache-2.0. It adds 21 tokens to every session and 1,842 once invoked, about $0.0001 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 skills, from other repositories