feat(#807): make code agent multi-forge (GitHub + GitLab)

.github/scripts/select-eval-agents-test.sh

@@ -96,7 +96,6 @@ YAML
   cat > "$tmpdir/harness/code.yaml" << 'YAML'
 agent: agents/code.md
 doc: docs/code.md
-policy: policies/code.yaml
 pre_script: scripts/pre-code.sh
 post_script: scripts/post-code.sh
 host_files:

FEATURES.md

@@ -162,7 +162,7 @@ agent prompt:
 If the new option requires the agent to reach a new external service
 from the sandbox:
 
-- [ ] Update `policies/<agent>.yaml` to allow the new host/port
+- [ ] Update `policies/<forge>/<agent>.yaml` to allow the new host/port
 - [ ] This is rare — most configuration changes don't need network
       changes
 

LOCAL.md

@@ -38,10 +38,18 @@ GitHub issue.
 Export the variables the agent needs:
 
 ```bash
+# GitHub:
 export GITHUB_ISSUE_URL="https://github.com/your-org/test-repo/issues/25"
 export GH_TOKEN="$(gh auth token)"
 export FULLSEND_FORGE="github"
 
+# GitLab (alternative — set these INSTEAD of the GitHub vars above;
+# only one forge's vars should be set at a time, and FULLSEND_FORGE
+# must match the chosen forge):
+# export GITLAB_ISSUE_URL="https://gitlab.com/your-group/test-project/-/issues/25"
+# export GITLAB_TOKEN="glpat-xxxxxxxxxxxxxxxxxxxx"
+# export FULLSEND_FORGE="gitlab"
+
 # GCP/Vertex AI credentials — required by most agents via
 # common/env/gcp-vertex.env and the host_files GOOGLE_APPLICATION_CREDENTIALS
 # mount in harness YAML.
@@ -56,9 +64,11 @@ If you're testing a new env var, export it here too. You can also use
 
 ### 2. Clone the target repo
 
-`GITHUB_ISSUE_URL` above points at an issue in a separate repo (e.g.
+The issue URL above points at an issue in a separate repo (e.g.
 `your-org/test-repo`) — clone it to its own local path so `--target-repo`
-has real content to work against:
+has real content to work against. The harness maps `GITHUB_ISSUE_URL` or
+`GITLAB_ISSUE_URL` to a generic `ISSUE_URL` via the per-forge env file
+(`env/github/*.env` or `env/gitlab/*.env`):
 
 ```bash
 git clone git@github.com:your-org/test-repo /tmp/target-repo

README.md

@@ -34,9 +34,9 @@ common/      Shared configuration (GCP Vertex AI auth)
 
 Agents run inside sandboxed containers with strict filesystem, network, and binary restrictions. Each agent follows a three-phase pipeline:
 
-1. **Pre-script** — runs on the GitHub Actions runner to validate inputs and prepare the environment
+1. **Pre-script** — runs on the CI runner to validate inputs and prepare the environment
 2. **Sandbox** — runs the agent with restricted permissions; the agent writes code and produces structured JSON output
-3. **Post-script** — runs on the runner with elevated permissions to perform GitHub mutations (pushing branches, creating PRs, posting comments, applying labels)
+3. **Post-script** — runs on the runner with elevated permissions to perform forge mutations (pushing branches, creating PRs/MRs, posting comments, applying labels)
 
 The agent never has direct write access to the repository. All mutations flow through post-scripts.
 

agents/code.md

@@ -1,7 +1,7 @@
 ---
 name: code
 description: >-
-  Implementation specialist for GitHub issues. Reads triaged issues, implements
+  Implementation specialist for issues. Reads triaged issues, implements
   fixes following repo conventions, runs tests and linters, and commits to a
   feature branch. Use when implementing a fix or feature from a triaged issue.
 model: opus
@@ -11,7 +11,7 @@ skills:
 
 # Code Agent
 
-You are an implementation specialist. Your purpose is to read a triaged GitHub
+You are an implementation specialist. Your purpose is to read a triaged
 issue, implement a fix or feature following the target repository's conventions,
 verify it passes tests and linters, and commit the result to a local feature
 branch. You do not triage issues, review PRs, push branches, create PRs, or

docs/code.md

@@ -2,7 +2,7 @@
 
 ![Code agent icon](icons/coder.png)
 
-Implementation specialist that reads triaged GitHub issues, implements fixes or features following repository conventions, runs tests and linters, and commits to a local feature branch.
+Implementation specialist that reads triaged issues, implements fixes or features following repository conventions, runs tests and linters, and commits to a local feature branch.
 
 ## Setup
 
@@ -45,8 +45,9 @@ See [Customizing with AGENTS.md](https://fullsend.sh/docs/guides/user/customizin
 
 | Variable | Description | Default | Valid values |
 |----------|-------------|---------|--------------|
-| `CODE_ALLOWED_TARGET_BRANCHES` | Restricts which branches the code agent can target when pushing. The post-code script validates the agent's chosen target branch against this variable before pushing. Set via `env.runner` in `harness/code.yaml` (never injected into the sandbox). | Repo default branch (auto-detected via GitHub API; falls back to `main`) | Comma-separated branch names (e.g. `main,develop`) or `*` for any branch |
-| `CODE_AUTO_MERGE` | Set to `"true"` to enable GitHub auto-merge on PRs created by the code agent. Requires branch protection with required reviews or status checks on the target branch. Read directly from the runner environment (not declared in `env.runner`). | `""` (disabled) | `"true"` to enable |
+| `CODE_ALLOWED_TARGET_BRANCHES` | Restricts which branches the code agent can target when pushing. The post-code script validates the agent's chosen target branch against this variable before pushing. Set via `env.runner` in `harness/code.yaml` (never injected into the sandbox). | Repo default branch (auto-detected via forge API; falls back to `main`) | Comma-separated branch names (e.g. `main,develop`) or `*` for any branch |
+| `FULLSEND_FORGE` | Forge platform. Set automatically by the harness `forge.<platform>.env` section. | (set by harness) | `"github"`, `"gitlab"` |
+| `CODE_AUTO_MERGE` | Set to `"true"` to enable auto-merge on PRs/MRs created by the code agent. On GitHub, uses `gh pr merge --auto`; on GitLab, uses `merge_when_pipeline_succeeds`. Requires branch protection with required reviews or status checks on the target branch. Read directly from the runner environment (not declared in `env.runner`). | `""` (disabled) | `"true"` to enable |
 | `CODE_AUTO_MERGE_METHOD` | Merge method for auto-merge: `"squash"`, `"rebase"`, or `"merge"`. When unset, auto-detected from the repo's allowed merge methods (prefers squash). Omitted automatically when the target branch uses a merge queue. Ignored unless `CODE_AUTO_MERGE` is `"true"`. | Auto-detected (prefers squash) | `"squash"`, `"rebase"`, `"merge"` |
 
 ## How the agent works
@@ -80,15 +81,16 @@ need a custom image.
 ### Image requirements
 
 A custom image must work within the constraints enforced by the sandbox
-policy ([`policies/code.yaml`](../policies/code.yaml)):
+policy ([`policies/github/code.yaml`](../policies/github/code.yaml) or
+[`policies/gitlab/code.yaml`](../policies/gitlab/code.yaml)):
 
 | Requirement | Detail |
 |-------------|--------|
 | **Base image** | Extend from `ghcr.io/fullsend-ai/fullsend-code:latest` to inherit the agent runtime, pre-installed tools, and security scanning binaries. |
 | **User/group** | The sandbox runs as `sandbox:sandbox`. Installed tools must be executable by this user. |
 | **Filesystem layout** | The working directory is `/sandbox/workspace`. Read-write access is limited to `/sandbox` and `/tmp`. System paths (`/usr`, `/lib`, `/etc`) are read-only at runtime — install packages at build time, not in an entrypoint. |
-| **Network access** | The sandbox restricts outbound network to specific hosts and binaries (Vertex AI, GitHub API, package registries). Arbitrary HTTP access is blocked. Tools that phone home at startup may fail. |
-| **Required binaries** | `git`, `gh`, `scan-secrets`, `pre-commit` must remain on `PATH`. Do not remove or shadow them. |
+| **Network access** | The sandbox restricts outbound network to specific hosts and binaries (Vertex AI, forge API, package registries). Arbitrary HTTP access is blocked. Tools that phone home at startup may fail. |
+| **Required binaries** | `git`, `scan-secrets`, `pre-commit` must remain on `PATH`. On GitHub, `gh` is also required; on GitLab, `curl` is used instead. Do not remove or shadow them. |
 
 ### How to build
 
@@ -110,7 +112,7 @@ RUN curl -fsSL https://deb.nodesource.com/setup_24.x | bash - \
 ```
 
 Build and push the image to a container registry accessible from your
-GitHub Actions runners:
+CI runners:
 
 ```bash
 docker build -t ghcr.io/<org>/<repo>-code:latest .
@@ -165,6 +167,28 @@ The precedence is as follows:
 **Note**: bots are filtered (`*[bot]`, `app/*`, `dependabot`). The resolution logic lives in
 [`scripts/lib/pr-assignee.lib.sh`](../scripts/lib/pr-assignee.lib.sh).
 
+## Multi-forge support
+
+The code agent supports both GitHub and GitLab. The harness
+`forge.<platform>` sections configure platform-specific policies,
+skills, env vars, and scripts. Key differences from single-forge
+setup:
+
+- **`FULLSEND_FORGE`** is required. Set automatically by the harness
+  `forge.<platform>.env` section (`"github"` or `"gitlab"`).
+- **`ISSUE_URL`** replaces `GITHUB_ISSUE_URL` in scripts. The
+  per-forge env file (`env/github/code.env` or `env/gitlab/code.env`)
+  maps the platform-specific variable to `ISSUE_URL`.
+- **Policy** is per-forge: `policies/github/code.yaml` or
+  `policies/gitlab/code.yaml`. Custom harnesses using `base:`
+  composition should override at the forge level if needed.
+- **GitLab uses `curl`** instead of `gh` for API access. The GitLab
+  sandbox policy allows `curl` for `gitlab_api` endpoints only.
+- **GitLab host allowlist** — `forge_validate_issue_url` in
+  `scripts/lib/gitlab-code-ops.lib.sh` and the network policy in
+  `policies/gitlab/code.yaml` both maintain an allowlist of GitLab
+  hosts. To support a self-hosted GitLab instance, update both files.
+
 ## Custom network policy
 
 If this agent needs to reach hosts beyond the defaults, see the

docs/network-policy.md

@@ -19,11 +19,11 @@ access with injected credentials. The GitLab forge policy allows `curl`
 because GitLab API access requires it (the `gh` CLI is not available).
 
 Each agent has its own default policy under
-[`policies/`](../policies/). The defaults cover Vertex AI, the GitHub
-API, package registries, and gitleaks releases; other agents have
-smaller subsets (for example, the scribe agent only needs Vertex AI).
-See the individual policy files for the full list of allowed hosts and
-binaries.
+[`policies/`](../policies/). The defaults cover Vertex AI, the forge
+API (GitHub or GitLab), package registries, and gitleaks releases;
+other agents have smaller subsets (for example, the scribe agent only
+needs Vertex AI). See the individual policy files for the full list of
+allowed hosts and binaries.
 
 ## Before you start
 
@@ -40,7 +40,7 @@ binaries.
 ### 1. Create a policy file
 
 Copy the default policy for the agent you are configuring (for example,
-[`policies/code.yaml`](../policies/code.yaml) for the code agent) to
+[`policies/github/code.yaml`](../policies/github/code.yaml) for the code agent) to
 `.fullsend/policies/<agent>.yaml` in your repository and append a
 block for the host you need:
 
@@ -97,8 +97,9 @@ Each agent has its own policy file with similar but not identical
 defaults. If multiple agents need the same custom hosts, create a
 separate override for each one. For example, the
 [code](code.md) and [fix](fix.md) agents use
-[`policies/code.yaml`](../policies/code.yaml) and
-[`policies/fix.yaml`](../policies/fix.yaml) respectively.
+[`policies/github/code.yaml`](../policies/github/code.yaml) and
+[`policies/fix.yaml`](../policies/fix.yaml) respectively
+(code agent also has a GitLab equivalent under `policies/gitlab/`).
 
 ## Troubleshooting
 

env/code.env

@@ -1,16 +1,7 @@
 export ISSUE_NUMBER=${ISSUE_NUMBER}
-export GITHUB_ISSUE_URL=${GITHUB_ISSUE_URL}
 
-# GH_TOKEN in the sandbox is a READ-ONLY scoped app installation token
-# (contents:read, issues:read, pull_requests:read). Set by
-# setup-agent-env.sh from CODE_GH_TOKEN. This token CANNOT push code
-# or create PRs — token scoping is the primary enforcement mechanism.
-# The separate write-enabled PUSH_TOKEN (env.runner) never enters the sandbox.
-export GH_TOKEN=${GH_TOKEN}
-
-# Git identity — uses the GitHub App bot user's noreply email so GitHub
-# links commits to the bot account (author.type === "Bot"). This makes
-# the Probot DCO app auto-exempt agent commits. The GIT_BOT_EMAIL var
+# Git identity — uses the bot user's noreply email so the platform
+# links commits to the bot account. The GIT_BOT_EMAIL var
 # is resolved at runtime by the "Resolve bot identity" workflow step.
 export GIT_AUTHOR_NAME="fullsend-code"
 export GIT_AUTHOR_EMAIL="${GIT_BOT_EMAIL}"

env/github/code.env

@@ -0,0 +1,3 @@
+export ISSUE_URL="${GITHUB_ISSUE_URL}"
+export GH_TOKEN=${GH_TOKEN}
+export FULLSEND_FORGE="github"

env/gitlab/code.env

@@ -0,0 +1,4 @@
+export ISSUE_URL="${GITLAB_ISSUE_URL}"
+export GITLAB_TOKEN="${GITLAB_TOKEN}"
+export GITLAB_HOST="${GITLAB_HOST}"
+export FULLSEND_FORGE="gitlab"

harness/code.yaml

@@ -4,17 +4,16 @@
 # Flow: pre_script → sandbox (agent) → post_script
 #   pre_script  : validates inputs on the runner BEFORE sandbox creation
 #   agent       : reads the issue, implements, tests, scans, commits locally
-#   post_script : secret scan, pre-commit, push branch, create PR, assign human
+#   post_script : secret scan, pre-commit, push branch, create PR/MR, assign human
 #
-# The agent NEVER pushes or creates PRs. Enforcement: the sandbox
-# GH_TOKEN is read-only scoped, PUSH_TOKEN never enters the sandbox,
-# and network policy restricts git/gh binary access.
+# The agent NEVER pushes or creates PRs/MRs. Enforcement: the sandbox
+# GH_TOKEN/GITLAB_TOKEN is read-only scoped, PUSH_TOKEN never enters the
+# sandbox, and network policy restricts binary access per forge.
 # Only the post-script, running on the runner with PUSH_TOKEN, can write.
 agent: agents/code.md
 doc: docs/code.md
 model: opus
 image: ghcr.io/fullsend-ai/fullsend-code@sha256:de3ecbd7719a1927c983142ada96475f3314d2505d0f258bcf19c31411856eb6
-policy: policies/code.yaml
 
 role: coder
 slug: fullsend-ai-coder
@@ -56,11 +55,49 @@ timeout_minutes: 35
 
 forge:
   github:
+    policy: policies/github/code.yaml
     pre_script: scripts/pre-code.sh
     post_script: scripts/post-code.sh
+    skills:
+      - skills/github-forge
+    host_files:
+      - src: env/github/code.env
+        dest: /sandbox/workspace/.env.d/forge-code.env
+        expand: true
     env:
       runner:
         PUSH_TOKEN: "${PUSH_TOKEN}"
         PUSH_TOKEN_SOURCE: "${PUSH_TOKEN_SOURCE}"
         REPO_FULL_NAME: "${REPO_FULL_NAME}"
         ISSUE_NUMBER: "${ISSUE_NUMBER}"
+        ISSUE_URL: "${GITHUB_ISSUE_URL}"
+        GH_TOKEN: "${GH_TOKEN}"
+        FULLSEND_FORGE: github
+      sandbox:
+        ISSUE_URL: "${GITHUB_ISSUE_URL}"
+        GH_TOKEN: "${GH_TOKEN}"
+        FULLSEND_FORGE: github
+  gitlab:
+    policy: policies/gitlab/code.yaml
+    pre_script: scripts/pre-code.sh
+    post_script: scripts/post-code.sh
+    skills:
+      - skills/gitlab-forge
+    host_files:
+      - src: env/gitlab/code.env
+        dest: /sandbox/workspace/.env.d/forge-code.env
+        expand: true
+    env:
+      runner:
+        PUSH_TOKEN: "${PUSH_TOKEN}"
+        PUSH_TOKEN_SOURCE: "${PUSH_TOKEN_SOURCE}"
+        REPO_FULL_NAME: "${REPO_FULL_NAME}"
+        ISSUE_NUMBER: "${ISSUE_NUMBER}"
+        ISSUE_URL: "${GITLAB_ISSUE_URL}"
+        GITLAB_TOKEN: "${GITLAB_TOKEN}"
+        GITLAB_HOST: "${GITLAB_HOST}"
+        FULLSEND_FORGE: gitlab
+      sandbox:
+        ISSUE_URL: "${GITLAB_ISSUE_URL}"
+        GITLAB_TOKEN: "${GITLAB_TOKEN}"
+        FULLSEND_FORGE: gitlab