AttackMap · Getting started

Quickstart

1. Scan a repository#

bash

attackmap analyze /path/to/repo --output reports

AttackMap auto-detects the languages in the repo, runs the relevant analyzers, and writes its results to reports/.

2. Read the results#

The most useful artifacts:

File	What it is
`reports/defensive-review.md`	Human-readable heuristic review
`reports/attackmap-report.json`	The full structured report (findings, surfaces, paths, exploitability)
`reports/*.md` sub-reports	Attack surface, attack paths, exploitability, diagrams

3. Add dependency CVEs#

bash

attackmap analyze /path/to/repo --output reports --cve

Cross-references your dependencies against OSV.dev and folds vulnerable packages into exploitability scoring (results cached for 24h).

4. Add an AI review#

bash

attackmap analyze /path/to/repo --output reports --llm

Generates a narrative defensive review with Claude (or OpenAI — see AI review). Every claim cites real evidence IDs, so the model can’t invent findings. Set an API key or sign in to the claude / codex CLI first.

5. Gate a pull request#

Compare against a baseline and fail CI if new high-severity findings appear:

bash
attackmap analyze . --baseline prev/attackmap-report.json \
  --diff-output reports/attackmap-diff.md --fail-on-new-high

See CI & pull requests for the GitHub Action and PR bot.

6. Scan a fleet#

Pass more than one repository to analyze a set of services together and surface the bugs that live in the seams between them:

bash

attackmap analyze ./service-a ./service-b ./gateway --output reports

Each repo gets its own reports/<repo>/, plus a reports/fleet-summary.md that links callers to the routes they hit and flags cross-boundary flows, trust-assumption gaps, and cross-repo anomalies. See Cross-repo fleet analysis.

7. Widen the net (recall)#

--recall widens taint discovery and enumerates capability-reach, marking the extra reach speculative; pair it with the verifier to adjudicate:

bash

attackmap analyze . --recall --hunt --verify

See Recall mode.

Edit this page on GitHub Report a problem