tamis

tamis does a first-pass triage of private vulnerability reports on GitHub. When a new security advisory arrives, an LLM agent investigates it against the repository and tamis puts the findings on a private project board.

For each report, tamis:

• lists exact duplicates, and names an embargoed one only once a separate check confirms it,
• finds public issues and pull requests that already fix or harden the same code,
• checks that the files and functions the report cites exist,
• reads the security policy and flags what may exclude the report,
• suggests a CWE and a CVSS 4.0 base vector.

The maintainers keep the decision. tamis only prepares it.

A triage looks like this on the board, with the cited code, the policy check and the severity folded in a details block. How it works shows the full layout.

### Automated triage

> Suggestions generated with an LLM. They help sort the report and do not replace a maintainer's review.

Scanned on 2026-10-24 at 14:32 UTC on `main` at `21a7561`.

Plausible signed `char` passed to `isxdigit()` in `php_url_decode_ex()`, the given reproducer never reaches it.

#### Duplicates

| Advisory | Status | Confidence |
|---|---|---|
| GHSA-xxxx-xxxx-xxxx (oldest report) | Published | High • Exact duplicate |

The agent runs in an isolated container driven by harness, the Alpha-Omega library behind scrutineer. Security describes what the sandbox allows.

flowchart LR
    subgraph gh["GitHub"]
        adv[("Advisory in triage")]
        others[("Other advisories,<br/>embargoed and closed")]
        pub["Public issues<br/>and pull requests"]
        board["Private project board"]
    end

    subgraph runner["Runner of the private workflow repository"]
        tamis["tamis"]
        proxy["Model proxy,<br/>holds the API key"]
        art[("Private artifact,<br/>kept 7 days")]
    end

    subgraph sandbox["Throwaway containers, no secret"]
        reader["Agent reading the report<br/>against a read-only clone"]
        dup["Offline agent<br/>searching duplicates"]
        confirm["Offline agent confirming<br/>one unpublished candidate"]
    end

    provider["Model provider"]

    adv -->|report and comments| tamis
    others -->|read with the App token| tamis
    board -->|"items already triaged<br/>and failed attempts"| tamis
    tamis -->|report and source| reader
    reader -->|unauthenticated search| pub
    reader -->|report.json| tamis
    pub -->|"existence, title and link<br/>of each cited issue"| tamis
    tamis -->|"report, checked clues,<br/>every other advisory"| dup
    dup -->|identifiers and confidence| tamis
    tamis -->|"candidate without<br/>its identifier"| confirm
    confirm -->|exact or not| tamis
    reader & dup & confirm -.->|model requests| proxy
    proxy --> provider
    tamis -->|"checked, escaped triage"| board
    tamis -->|complete result| art

tamis sends the report to the model provider you choose. Check that your disclosure policy allows it.

tamis init runs these three steps for you, all but installing the App, as the setup guide shows. By hand:

1. Create a GitHub App for your organisation with the "Repository security advisories" permission and the organisation "Projects" permission, both in read and write. Install it on the project repository only.
2. In a private repository, store the App's client ID and private key, and an Anthropic or OpenAI API key or a Codex access token.
3. Add this workflow to the private repository:

name: tamis

on:
  schedule:
    - cron: "*/15 * * * *"
  workflow_dispatch:
    inputs:
      advisory:
        description: GHSA to triage
        required: false

concurrency:
  group: tamis

jobs:
  triage:
    runs-on: ubuntu-latest
    timeout-minutes: 240
    environment: tamis
    permissions:
      contents: read
    steps:
      - uses: actions/create-github-app-token@64ab729a6294000f5aa4a4bc0b73d1f32bc45510 # v3.3.0
        id: app
        with:
          client-id: ${{ vars.TAMIS_CLIENT_ID }}
          private-key: ${{ secrets.TAMIS_PRIVATE_KEY }}
          owner: your-org
          repositories: your-project
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          repository: your-org/your-project
          fetch-depth: 0
          persist-credentials: false
      # Replace this placeholder with the full SHA of a reviewed tamis release, and its tag.
      - uses: php/tamis@REPLACE_WITH_REVIEWED_FULL_COMMIT_SHA # vYYYY.MM.DD.N
        with:
          repository: your-org/your-project
          # A date, or an RFC 3339 time with its offset: "2026-10-22T14:30:00+02:00"
          since: "2026-10-22"
          advisory: ${{ inputs.advisory }}
          api-key: ${{ secrets.ANTHROPIC_API_KEY }}
          github-token: ${{ steps.app.outputs.token }}
          bot-login: your-app-slug[bot]
          # List every branch your project supports. Each ref must exist in the checkout.
          references: master origin/release-branch
          # The private board the triages go on, created with `tamis create-board your-org`.
          project: your-org/NUMBER

Pin the commit of a release tag, since the action then runs the attested binary of that release instead of building tamis from source. Set since to the installation date so tamis leaves existing advisories alone. Keep the workflow repository private, since the action uploads the reports there as artifacts kept for seven days. The setup guide covers source branches and provider spending limits.

MIT