company-codegraph

A Codex skill for exploring a codebase's structure through an approved local code graph, then checking important claims against source files and tests.

In plain words
What is it for?
Use it to find symbols, callers, dependencies, architecture boundaries, and possible files affected by a change.
Why use it?
It helps answer structural code questions while keeping graph access local, read-only, and bounded.

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/r1ck29/claude-code-codegraph-harness/company-codegraph
Any agent
npx skills add R1ck29/claude-code-codegraph-harness --skill company-codegraph
Clone the repo
git clone --depth 1 https://github.com/R1ck29/claude-code-codegraph-harness

Made for: Claude Code, Codex.

Per session 29 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 556 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 $0.00029 $0.00556
Opus 5 $0.00015 $0.00278
Sonnet 5 $0.00006 $0.00111
Haiku 4.5 $0.00003 $0.00056

Measured yesterday against content hash 986c183ceed7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

company-codegraph 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.

codex/skills/company-codegraph/SKILL.md · 60 lines

How it starts

The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Company codegraph for Codex

This file is the distributable source for the user skill installed at $HOME/.agents/skills/company-codegraph/SKILL.md. It does not modify a global or project AGENTS.md.

Apply this skill only when the question is structural. The configured MCP server is a local gateway; this skill does not register the server and does not prove that its index is usable.

Data boundary

Use only the registered local gateway. Do not register or call an upstream graph backend directly. Do not send source or graph output to another service.

Allowed graph tools

Only these five read-only gateway tools are in scope:

  • codegraph_status
  • codegraph_search
  • codegraph_neighbors
  • codegraph_impact
  • codegraph_architecture

Every normal query revalidates freshness and returns it in the same response. Inspect that field; do not add a redundant codegraph_status preflight. Use codegraph_status only for diagnostics when no structural query is otherwise needed. Keep search limits and traversal depth bounded.

Routing

Use the graph only for structural tasks:

  • symbol discovery
  • callers and callees
  • dependency paths
  • neighbors and hubs
  • architecture boundaries
  • change-impact candidates

Use Read, Grep, LSP and relevant tests first for:

  • exact control flow
  • error and validation behavior
  • runtime dependency injection
  • reflection and generated code
  • configuration-driven behavior
  • final change decisions

Fail-closed fallback

If status or any result is unavailable, stale, dirty, truncated, stop using graph evidence for the task and fall back to Read, Grep, LSP. Do not ask the user to repair graph state during a coding task.

Evidence boundary

The graph is derived static-analysis data, not source-of-truth. The current source and tests are the source-of-truth. Open consequential source locations and verify relevant tests before proposing or making a change. Clearly separate graph-derived candidates from claims verified in source and tests.

Read the full file on GitHub · 60 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 · 60 lines · 29 tokens per session scan A 986c183ceed7

Subscribe to this mod's changes

company-codegraph is a skill published in the GitHub repository R1ck29/claude-code-codegraph-harness (0 stars, last pushed 10d ago), licensed Apache-2.0. It adds 29 tokens to every session and 556 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 2026-08-31.

Related

Other skills, from other repositories

sem

Use sem to get entity-level (function/class/method) semantic diffs, impact analysis, blame, and dependency context from any Git repo. Trigger this skill whenever the user asks what changed in a commit or PR, wants to understand the blast radius of a change, needs to know who last modified a function, wants to trace…

Ataraxy-Labs/sem · 112 tokens

octocode-benchmark

Use when planning, running, grading, or reporting the by-hand Octocode research benchmark — pairwise matchups (Octocode anchor vs one baseline: gh+RTK, gh+Headroom, or plain gh) over markdown questions, with a fresh isolated runner agent per (question, arm, pass), one blind judge per question grading two answers X/Y…

bgauryy/octocode · 117 tokens

x-ray

Deep codebase exploration using semantic search and relationship mapping. Use when you need to understand the current codebase.

bartolli/codanna · 25 tokens

roam

Codebase comprehension via roam-code CLI. Use when exploring codebases, planning modifications, debugging failures, assessing PR risk, or checking architecture health. Triggers on: understanding project structure, pre-change safety checks, finding symbols/files, blast radius analysis, affected tests, health scoring…

Cranot/roam-code · 86 tokens

octocode-roast

Use when code needs a blunt evidence-backed roast or memorable critique: smell inventory, debt ranking, hot-path autopsy, savage/diff review, security or performance sins, or practical redemption paths. Phrases like roast this, brutal review, top sins, cleanup debt. Polite evidence-first PR review → octocode-research.

bgauryy/octocode · 71 tokens

boutique-best-practices

Best practices for using Boutique with Swift 6 concurrency, @Observable, @ObservationIgnored, Sendable conformance, testing with preview stores, and dependency injection. Use when troubleshooting Boutique issues, migrating to Swift 6, or setting up tests.

mergesort/Boutique · 56 tokens