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.
npx skills add DiUS/agent-toolkit --skill codebase-discoverygit clone --depth 1 https://github.com/DiUS/agent-toolkitWrote 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.
[](https://agentmods.dev/skills/dius/agent-toolkit/codebase-discovery)<a href="https://agentmods.dev/skills/dius/agent-toolkit/codebase-discovery"><img src="https://agentmods.dev/badge/skills/dius/agent-toolkit/codebase-discovery/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.
<a href="https://agentmods.dev/skills/dius/agent-toolkit/codebase-discovery"><img src="https://agentmods.dev/badge/skills/dius/agent-toolkit/codebase-discovery.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00107 | $0.04837 |
| Opus 5.5 | $0.00043 | $0.01935 |
| Sonnet 5.5 | $0.00021 | $0.00967 |
| Haiku 4.5 | $0.00011 | $0.00484 |
Grade A, and why
codebase-discovery 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 2mo 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.
How it starts
The opening of the file, as written. The whole thing — 358 lines — stays where its author put it; the contents beside it link to each section on GitHub.
User Input
$ARGUMENTS
You MUST consider the user input before proceeding (if not empty). It says what to analyse and how, either in plain words or with the options below.
Options
Each one pre-answers a question this skill would otherwise ask, or overrides something it would infer. All optional; absent means work it out as usual.
| Argument | Effect |
|---|---|
full | code-only |
the mode (see Modes) |
--exclude <globs> |
additional exclusions, gitignore syntax — see recon-heuristics |
--output <dir> |
the output root, instead of agreeing it in Phase 0 |
--fresh |
start cold instead of resuming. Where a previous run exists, Phase 0 confirms first — a clean run discards its _discovery/ files (see Phase 0) |
--on-drift <recon|full-recon|proceed|report> |
pre-answer the freshness check's question |
--interview |
enter at Phase 2 and continue the interview queue |
How to read them
- Resolve the request, however it's phrased. These are a shorthand, not the interface: "skip the test projects, don't touch the docs site, and there's nobody to interview yet" must land on the same settings as the equivalent flags. Extract from prose, flags, or a mix.
- Echo the resolved set back in one line before starting — "code-only ·
excluding
tests/*· outputdocs/discovery/· drift → recon". There's no parser; the user needs to see what was understood. - Report anything you couldn't resolve; never guess. A silently dropped
--excludemeans reading a tree the user told you to leave alone, and "skip the old stuff" needs a question, not a decision. - An option means don't ask that question — state the value you were given and move on.
- But no option authorises discarding existing work. It pre-answers a choice, not a deletion, so
--freshover a previous run still needs sign-off (see Phase 0). - Record the resolved options in
discovery-state.md, so a resumed run reuses them.
What ships with it
30 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- playbooks/00-pre-check.md 12 KB
- playbooks/01-deep-recon.md 12 KB
- playbooks/02-interview.md 10 KB
- playbooks/03-synthesis.md 7.1 KB
- playbooks/04-verification.md 9.1 KB
- playbooks/05-finish.md 4.5 KB
- README.md 8.6 KB
- references/code-intelligence.md 6.0 KB
- references/discovery-disposition.md 2.4 KB
- references/freshness.md 3.5 KB
- references/navigation.md 6.0 KB
- references/output-conventions.md 14 KB
- references/provenance-and-status.md 11 KB
- references/question-bank.md 3.4 KB
- references/recon-heuristics.md 5.7 KB
- references/write-contract.md 3.3 KB
- templates/agent-onboarding-file.md 1.8 KB
- templates/assumptions-register.md 1.7 KB
- templates/business-requirements.md 1.6 KB
- templates/business-rules.md 1.7 KB
- templates/current-architecture.md 2.2 KB
- templates/discovery-state.md 3.3 KB
- templates/domain-glossary.md 1.0 KB
- templates/domain-model.md 2.0 KB
- templates/integrations.md 1.2 KB
- templates/project-readme.md 2.9 KB
- templates/recon-manifest.md 2.8 KB
- templates/traceability-index.md 981 B
- templates/user-personas.md 876 B
- templates/workflows.md 1.4 KB
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.
- 2mo ago First seen · 358 lines · 107 tokens per session scan A a37a9a5b3cff
codebase-discovery is a skill published in the GitHub repository DiUS/agent-toolkit (2 stars, last pushed 21d ago), licensed MIT. It adds 107 tokens to every session and 4,837 once invoked, about $0.0004 per session on Opus 5.5. 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.
Other skills, from other repositories
connect
Install the scoped Antigravity permission hook.
hook-template
Generate hook script from template. Use when adding a new hook, wiring a PreToolUse/PostToolUse/Stop/Notification hook, or scaffolding hook config for settings.json.
check-environment
Verify Claude, Codex, and Grok availability plus Director guidance, relay, agents, and skills. Audit optional hooks only when selected. Use after installation or when a native surface misbehaves.
test-runner
Test execution reference: framework detection (pytest/jest/vitest/go/cargo/junit) and correct run/coverage commands, plus failure-analysis steps. Use when running tests, analyzing test failures, or verifying coverage after code changes.
agents
List all available agents (core, expert, self-evolving). Use when the user asks what agents are available or runs /agents.
handoff-gemini
Delegate long-context analysis, research, and large-document summarization to Google Gemini CLI (1M-token context) via non-interactive gemini -p, preserving Claude context. Use when the user says 'use gemini'/'hand off to gemini' or a task needs 100K+ tokens of context.