docs: propose labels for Jenkins/GlitchTip workflow (#80)

Merge PR 87

Co-authored-by: Jason Walker <913443@dadeschools.net>
Co-committed-by: Jason Walker <913443@dadeschools.net>

docs/architecture/jenkins-readonly-build-status-design.md

@@ -0,0 +1,151 @@
+# Jenkins Read-Only Build Status Tools — Design Notes
+
+- **Status:** Design (implementation-ready notes; **no implementation in this repo**)
+- **Issue:** #72 (parent umbrella: #75; boundary decision: ADR-0001, #71)
+- **Related:** #77 (repo/branch/PR → job mapping, designed separately)
+- **Date:** 2026-07-02
+
+## 1. Purpose and scope
+
+Define the minimum **read-only** Jenkins MCP tool set that lets an LLM answer:
+*"Did the latest build for this project/branch succeed or fail?"* — plus enough
+detail (build URL, number, timing, result) to report or investigate.
+
+Phase 1 is **strictly read-only**, per ADR-0001
+([`adr-0001-mcp-control-plane-boundaries.md`](adr-0001-mcp-control-plane-boundaries.md)):
+
+- **Excluded: build triggers.**
+- **Excluded: deploy triggers.**
+- **Excluded: parameterized job launches.**
+- Excluded: job creation/deletion/config changes, queue manipulation, node
+  management — any Jenkins mutation whatsoever.
+
+## 2. Boundary placement
+
+These tools belong to the **`jenkins-mcp`** package/server of the MCP Control
+Plane — **never** inside `gitea-mcp` (`mcp_server.py` in this repo).
+Consequences (from `tool-boundaries.md`, `credential-isolation.md`, ADR-0001):
+
+- `jenkins-mcp` runs as its own server process with its own `.env`.
+- **Jenkins credentials never enter the Gitea MCP runtime**, and Gitea
+  credentials never enter `jenkins-mcp`.
+- This document lands in this repo only because the repo currently hosts the
+  Control Plane's architecture docs; the code ships elsewhere (owner decision
+  #1 of ADR-0001).
+
+## 3. Minimum read-only tool set
+
+| Tool | Purpose |
+|---|---|
+| `jenkins_whoami` | Verify authenticated Jenkins identity + active profile (mirror of `gitea_whoami`; fail-closed identity proof before anything else) |
+| `jenkins_list_jobs` | List visible jobs (supports folder paths), with pagination bounds |
+| `jenkins_latest_build` | The primary question: latest build of a job (or job+branch for multibranch) → status summary |
+| `jenkins_build_status` | Status of a specific build number (job, number) |
+| `jenkins_get_build` | Full safe detail of a build (fields in §4) |
+| `jenkins_console_tail` | Bounded, redacted tail of a build's console log (§6) — optional, approval-gated addition |
+
+All tools are `GET`-only against the Jenkins JSON API (`/api/json`,
+`.../lastBuild/api/json`, `.../consoleText`). No tool issues POST/PUT/DELETE.
+
+## 4. Return payloads (safe fields)
+
+`jenkins_latest_build` / `jenkins_build_status` / `jenkins_get_build` return:
+
+| Field | Source | Notes |
+|---|---|---|
+| `job` | request | Fully-qualified job path (folders joined with `/`) |
+| `build_number` | `number` | int |
+| `result` | `result` | `SUCCESS` / `FAILURE` / `UNSTABLE` / `ABORTED` / `NOT_BUILT`; `null` → `IN_PROGRESS` when `building=true` |
+| `building` | `building` | bool |
+| `url` | `url` | Build URL |
+| `branch` | multibranch job name / SCM action | Best-effort; omitted when unknown |
+| `timestamp` | `timestamp` | ISO-8601 UTC (converted from epoch ms) |
+| `duration_seconds` | `duration` | 0/omitted while building |
+| `commit_sha` | SCM build action | Best-effort; omitted when unknown |
+
+Rules: no raw Jenkins payload passthrough (allowlist projection only); no
+`Authorization` header, token, or crumb material in any output or error
+(reuse the shared redaction approach of `safety-model.md` §3 / `gitea_audit`).
+
+## 5. Failure behavior (fail closed, clear, safe)
+
+| Condition | Behavior |
+|---|---|
+| Unknown job | Explicit `{"found": false, "job": "<path>", "error": "job not found"}` — never guess or fuzzy-match a job name (hard rule; see also #77) |
+| Jenkins unreachable (DNS/timeout/conn refused) | Clear `"network error contacting Jenkins: <redacted reason>"`; no retry storm — mirror `gitea_auth.api_request` timeout + failure conversion |
+| 502/503/504 | Explicit "Jenkins upstream unavailable" |
+| 401/403 | "Jenkins auth failed / insufficient permissions" — **without** echoing credentials or the request's auth material |
+| Malformed JSON | "malformed JSON response from Jenkins" (no raw-body dump) |
+| Missing profile/creds | Fail closed before any network call (§7) |
+
+## 6. Console tail safety (`jenkins_console_tail`)
+
+Console logs are the highest-risk surface (secrets, tokens, internal hosts
+routinely leak into build logs). If included at all (owner may defer it):
+
+- **Bounded:** hard server-side cap (default: last 200 lines AND ≤ 64 KiB,
+  whichever is smaller; caller may request less, never more).
+- **Redacted:** pass through the shared secret redactor (token/`Basic`/`Bearer`/
+  password/key-value patterns) before returning; redaction failure ⇒ return an
+  error, never the raw text.
+- **Default off:** summary fields (`result`, failing stage if cheaply available)
+  are preferred; the tail requires an explicit `allowed_operations` entry
+  (`jenkins.console.read`) distinct from plain `jenkins.build.read`.
+
+## 7. Credentials and profile requirements
+
+Follows the per-service profile model (`gitea-execution-profiles.md`, extended
+by #76):
+
+- Env/config: `JENKINS_URL`, `JENKINS_USER`, `JENKINS_TOKEN_SOURCE_NAME`
+  (name-of-secret only — value resolved at runtime, never logged/committed).
+- Profile: e.g. `jenkins-readonly` with namespaced
+  `allowed_operations: ["jenkins.read", "jenkins.build.read"]`
+  (+ `jenkins.console.read` only if the tail tool is approved);
+  `forbidden_operations: ["jenkins.build.trigger", "jenkins.deploy", "jenkins.job.configure"]`
+  as belt-and-braces even though no mutating tool exists.
+- Missing URL/user/token/profile ⇒ **fail closed** with a clear message.
+- Since every tool is read-only, no confirmation gates are needed — but
+  identity (`jenkins_whoami`) must still work so workflows can prove which
+  Jenkins account they act as.
+
+## 8. Job addressing and mapping
+
+Tools accept an explicit fully-qualified job path (folder-aware:
+`folder/subfolder/job`). How a *repo/branch/PR* resolves to that job path is
+**out of scope here** and designed in **#77**, with these fixed constraints:
+
+- No silent guessing of job names — unmapped input returns an explicit
+  "no mapping" result.
+- Multibranch pipelines address a branch job as `<job>/<branch>` with proper
+  URL-encoding of branch names (e.g. `feature%2Fx`).
+
+## 9. Testing strategy (for the implementing package)
+
+Mocked-Jenkins unit tests only (no live Jenkins in unit CI), mirroring this
+repo's conventions (`docs/developer-testing-guidelines.md`):
+
+- Patch the HTTP client; assert method is always `GET` and URL shape is correct
+  (folders, multibranch encoding).
+- Success projections: field allowlist exactly as §4; unknown fields dropped.
+- `result=null + building=true` ⇒ `IN_PROGRESS`.
+- Unknown job ⇒ found:false, no fuzzy match, no API retry.
+- Timeout/DNS/5xx/malformed-JSON ⇒ safe errors, no secret/credential leakage
+  (explicit no-token-in-error assertions).
+- Console tail: cap enforcement (lines and bytes), redaction applied, redaction
+  failure ⇒ error not raw text, gated behind `jenkins.console.read`.
+- Profile gate: missing/insufficient profile ⇒ no network call
+  (`mock_api.assert_not_called()` pattern).
+
+## 10. Implementation-readiness checklist
+
+Ready to implement in `jenkins-mcp` once:
+
+1. ADR-0001 owner decision #1 (where `jenkins-mcp` lives) is made.
+2. #76 profile schema exists (or a minimal `jenkins-readonly` profile is
+   hand-rolled to the same rules).
+3. #77 mapping design is accepted (or tools ship path-addressed only, mapping
+   deferred).
+
+Explicitly **not** unlocked by this document: build triggers, deploys,
+parameterized launches, any Jenkins code in `mcp_server.py`.

docs/label-taxonomy.md

@@ -0,0 +1,34 @@
+# Label Taxonomy
+
+This document catalogs the issue labels used for MCP workflows, including Jenkins and GlitchTip (observability).
+
+> **Approval Required:** Do not create or apply new labels in `manage_labels.py` without explicit owner approval of this document.
+
+## Existing Labels
+
+* **`jenkins`**
+  * Description: Jenkins integration
+  * Color: `d93f0b`
+  * Use: Used to mark issues, PRs, or tasks that involve the `jenkins-mcp` boundaries, CI/CD designs, or build failures.
+
+* **`glitchtip`**
+  * Description: GlitchTip integration
+  * Color: `b60205`
+  * Use: Used to mark issues related to the `glitchtip-mcp` boundary and observability integration.
+
+## Proposed / Missing Labels
+
+* **`observability`**
+  * Proposed Description: Observability, metrics, and monitoring tasks
+  * Proposed Color: `5319e7`
+  * Use: Broader than GlitchTip alone; covers logging, metrics, traces, and general observability pipeline improvements.
+
+* **`source:glitchtip`**
+  * Proposed Description: Issue filed automatically by GlitchTip orchestration
+  * Proposed Color: `b60205`
+  * Use: Applied automatically by the orchestrator when a GlitchTip error event is converted into a Gitea issue.
+
+* **`status:triage`**
+  * Proposed Description: Issue needs human or orchestrator triage
+  * Proposed Color: `fbca04`
+  * Use: Used for incoming issues (especially automated ones like `source:glitchtip`) that have not yet been evaluated for priority or resolution.

docs/llm-agent-sha.md

@@ -0,0 +1,138 @@
+# LLM-Agent-SHA — Opaque Agent Attribution (Phase 0)
+
+Convention for attributing work to a specific LLM session/workstream across
+issues, branches, PRs, and review handoffs, without exposing a human or model
+identity. Approved by the owner decision on issue #86
+(`#issuecomment-1354`); this document implements **Phase 0 only**.
+
+## The one rule that matters
+
+`LLM-Agent-SHA` is **informational attribution metadata only**. It must never
+be used for authentication, authorization, review eligibility, merge
+eligibility, profile permissions, or any other security decision.
+
+The security gates remain, unchanged:
+
+- the **authenticated Gitea user** (self-review/self-merge protection),
+- the **active MCP profile** and its `allowed_operations`
+  (see [`gitea-execution-profiles.md`](gitea-execution-profiles.md)),
+- the fail-closed eligibility checks in `gitea_check_pr_eligibility`.
+
+Two sessions with different `LLM-Agent-SHA` values that authenticate as the
+same Gitea user are **the same actor** for review/merge safety. A different
+SHA never unlocks self-review or self-merge. `tests/test_llm_agent_sha.py`
+proves the eligibility logic never consults the SHA.
+
+## Format
+
+```text
+LLM-Agent-SHA: llm-<12 lowercase hex chars>
+```
+
+Validation regex:
+
+```text
+^llm-[0-9a-f]{12}$
+```
+
+Examples: `llm-8f3a9c2d6b41`, `llm-41d0e7aa9f2c`, `llm-b7c93d441a08`.
+
+### Generation
+
+Generate 48 random bits, e.g. `python3 -c "import secrets; print('llm-' +
+secrets.token_hex(6))"`, or hash a non-secret session UUID. An
+operator-provided opaque ID is also fine.
+
+Do **not** derive the value from any of:
+
+- a Gitea token or other secret,
+- an email address or username,
+- a machine hostname or private filesystem path,
+- a model or provider name,
+- conversation contents.
+
+The SHA must contain no model name, provider name, human name, email,
+hostname, token, private path, or conversation-derived content. It is safe to
+include in PR bodies, issue comments, and audit logs — and only there.
+
+## Lifetime
+
+Canonical lifetime is **per PR/workstream**: pick one SHA when starting an
+issue and keep it through the branch, PR, and handoff for that workstream. A
+per-session SHA is acceptable when the session maps cleanly to one
+workstream. Do not reuse a SHA across unrelated workstreams.
+
+## Placement
+
+Phase 0 uses **visible markdown metadata blocks** (not hidden HTML
+comments). Include the block in PR bodies and review handoffs; keep it out of
+ordinary comments unless attribution is genuinely useful there.
+
+**Never put the SHA in branch or worktree names.** Branches stay
+issue-linked and human-readable (`docs/issue-86-llm-agent-sha-phase0`), per
+the branch standard.
+
+### Handoff metadata block (implementer → PR body / handoff report)
+
+```markdown
+LLM Handoff Metadata:
+- LLM-Agent-SHA: llm-8f3a9c2d6b41
+- LLM-Role: implementer
+- Authenticated-Gitea-User: jcwalker3
+- MCP-Profile: gitea-default
+- Branch: docs/example-branch
+- Worktree: branches/docs-example-branch
+- Self-review allowed: no
+```
+
+### Review metadata block (reviewer → review comment)
+
+```markdown
+Review Metadata:
+- LLM-Agent-SHA: llm-41d0e7aa9f2c
+- LLM-Role: reviewer
+- Authenticated-Gitea-User: sysadmin
+- MCP-Profile: prgs-reviewer
+- Eligibility: passed
+```
+
+## Same SHA vs same user vs same profile
+
+Reviewers and operators must keep three distinct identities straight:
+
+| Comparison | Meaning | Effect on eligibility |
+|---|---|---|
+| same `LLM-Agent-SHA` | same LLM session/workstream wrote both artifacts | **none — attribution only** |
+| same authenticated Gitea user | same Gitea actor | **blocks** self-review / self-merge, regardless of SHA |
+| same MCP profile | same capability set | governs `allowed_operations` (what actions are permitted at all) |
+
+Concretely: an implementer session (`llm-8f3a…`, user `jcwalker3`) and a
+would-be reviewer session (`llm-41d0…`, also user `jcwalker3`) have different
+SHAs but the **same Gitea user** — the reviewer session is still the PR
+author to Gitea and must not review, approve, or merge. Review handoffs
+require a genuinely different authenticated user (e.g. `sysadmin` /
+`prgs-reviewer`).
+
+## Phase 0 scope (and what is deferred)
+
+Phase 0 is documentation, handoff/review templates, and negative tests only.
+Deferred to later owner-approved phases; none of this exists yet:
+
+- launcher-enforced SHA generation,
+- `LLM_AGENT_SHA` / `LLM_AGENT_ROLE` environment injection,
+- `gitea_whoami` returning SHA/role,
+- automatic PR body injection by MCP tools,
+- audit schema changes requiring the SHA,
+- release/orchestrator lineage tracking.
+
+MCP tools neither read nor emit the SHA. Setting an `LLM_AGENT_SHA`
+environment variable has no effect on any tool; the negative tests assert
+eligibility results are byte-identical with and without it.
+
+## Related documents
+
+- [`llm-workflow-runbooks.md`](llm-workflow-runbooks.md) — the runbooks whose
+  handoffs carry these blocks
+- [`gitea-execution-profiles.md`](gitea-execution-profiles.md) — profiles and
+  `allowed_operations` (the real permission gate)
+- [`safety-model.md`](safety-model.md) — audit, redaction, confirmation gates

docs/llm-workflow-runbooks.md

@@ -45,6 +45,18 @@ Use any eligible reviewer profile to review PR #N.
 Use any eligible merger profile to merge PR #N if checks pass.
 ```
 
+### Attribution: `LLM-Agent-SHA` (metadata only)
+
+Sessions may attribute their work with an opaque `LLM-Agent-SHA`
+(`llm-<12 lowercase hex>`, e.g. `llm-8f3a9c2d6b41`) in PR-body and
+review-handoff metadata blocks — see
+[`llm-agent-sha.md`](llm-agent-sha.md) for the full convention. It is
+**attribution only**: eligibility is decided solely by the authenticated
+Gitea user and the profile's allowed operations. Two sessions with different
+SHAs under the same Gitea user are the same actor — a different SHA never
+permits self-review or self-merge. Keep the SHA out of branch and worktree
+names.
+
 ## Prerequisites: canonical config + thin launchers
 
 Runtime profiles live in **one canonical JSON file**, referenced by every LLM
@@ -274,7 +286,8 @@ touching anything.
   `fix/...` / `docs/...`); `cd` into that worktree; implement narrowly; add or
   update tests if behavior changes; run the full suite; commit with an
   issue-linked message; open a PR to `master`. **Do not** review or merge your
-  own PR.
+  own PR. Include an `LLM Handoff Metadata` block (with `LLM-Agent-SHA`) in
+  the PR body — see [`llm-agent-sha.md`](llm-agent-sha.md).
 - **Prompt:** `Use an author profile to implement issue #N and open a PR to
   master. Do not self-review or self-merge.`
 
@@ -285,7 +298,11 @@ touching anything.
 - **Steps:** confirm identity + eligibility (menu eligibility check or
   `gitea_check_pr_eligibility`); read the diff; confirm scope matches the linked
   issue; post the review (`comment` / `request_changes` / `approve`) via the
-  gated review tool. Pin the reviewed head SHA where supported.
+  gated review tool. Pin the reviewed head SHA where supported. Include a
+  `Review Metadata` block (with your own `LLM-Agent-SHA`) in the review —
+  and remember: a different `LLM-Agent-SHA` does **not** make you a different
+  actor; only a different authenticated Gitea user does
+  ([`llm-agent-sha.md`](llm-agent-sha.md)).
 - **Prompt:** `Use any eligible reviewer profile to review PR #N. Approve only
   if scope matches issue #M and checks pass; otherwise request changes.`
 
@@ -391,6 +408,7 @@ scripts/release-tag v0.4.0 --notes-file /tmp/release-notes.md --push
 
 - [`../skills/llm-project-workflow/SKILL.md`](../skills/llm-project-workflow/SKILL.md) — portable cross-project LLM workflow skill.
 - [`gitea-execution-profiles.md`](gitea-execution-profiles.md) — the profile model.
+- [`llm-agent-sha.md`](llm-agent-sha.md) — opaque agent attribution metadata (never an eligibility input).
 - [`safety-model.md`](safety-model.md) — trust boundaries and audit logging.
 - [`tool-boundaries.md`](tool-boundaries.md) — per-tool allowed operations.
 - [`credential-isolation.md`](credential-isolation.md) — credential handling.

skills/llm-project-workflow/templates/review-pr.md

@@ -8,6 +8,8 @@ Task: review PR #<pr> for issue #<n>.
 Rules (llm-project-workflow):
 - Review in a SEPARATE detached review worktree, never the author's folder.
 - You must NOT be the PR author. If the authenticated user == PR author, stop.
+  A different LLM-Agent-SHA does NOT make you a different actor — only a
+  different authenticated Gitea user does (docs/llm-agent-sha.md).
 - Do not merge if any check fails.
 
 Steps:
@@ -21,6 +23,14 @@ Steps:
 6. Run the test suite; note results.
 7. Post the review verdict: approve only if scope is clean and checks pass;
    otherwise request changes with specifics. Never merge from this review step.
+   Include a "Review Metadata" block (attribution only — docs/llm-agent-sha.md):
+
+   Review Metadata:
+   - LLM-Agent-SHA: llm-<12 lowercase hex, e.g. llm-41d0e7aa9f2c>
+   - LLM-Role: reviewer
+   - Authenticated-Gitea-User: <whoami result>
+   - MCP-Profile: <profile name>
+   - Eligibility: passed/failed
 
 Handoff: reviewer identity, PR author, scope verdict, checks + results, decision.
 ```

skills/llm-project-workflow/templates/start-issue.md

@@ -23,6 +23,17 @@ Steps:
 6. Checks: run the test suite, compile/lint changed files, git diff --check,
    and scan the diff for secrets.
 7. Commit (issue-linked message), push the branch, open a PR to master.
+   Include an "LLM Handoff Metadata" block in the PR body (attribution only;
+   never an eligibility input — docs/llm-agent-sha.md):
+
+   LLM Handoff Metadata:
+   - LLM-Agent-SHA: llm-<12 lowercase hex, e.g. llm-8f3a9c2d6b41>
+   - LLM-Role: implementer
+   - Authenticated-Gitea-User: <whoami result>
+   - MCP-Profile: <profile name>
+   - Branch: <branch>
+   - Worktree: <worktree path>
+   - Self-review allowed: no
 8. Stop before review/merge — you are the author.
 
 Handoff: issue #, branch, worktree path, files changed, checks + results, PR URL.

tests/test_llm_agent_sha.py

@@ -0,0 +1,195 @@
+"""Negative tests for LLM-Agent-SHA attribution (#86, Phase 0).
+
+``LLM-Agent-SHA`` (docs/llm-agent-sha.md) is attribution metadata ONLY. These
+tests prove it can never bypass the review/merge safety gates:
+
+1. Same Gitea user + different LLM-Agent-SHA still fails self-review/approval.
+2. Same Gitea user + different LLM-Agent-SHA still fails self-merge.
+3. The eligibility logic does not consult SHA metadata at all — results are
+   identical with no SHA, one SHA, or a different SHA in the environment, and
+   no gate accepts an agent-SHA input.
+
+Phase 0 adds no SHA support to any MCP tool; the environment variables set
+here (``LLM_AGENT_SHA`` / ``LLM_AGENT_ROLE``) simulate a future launcher and
+must be ignored by every gate.
+"""
+import inspect
+import os
+import re
+import sys
+import unittest
+from unittest.mock import patch
+
+sys.path.insert(0, str(__import__("pathlib").Path(__file__).resolve().parent.parent))
+
+from mcp_server import (  # noqa: E402
+    gitea_check_pr_eligibility,
+    gitea_merge_pr,
+    gitea_review_pr,
+    gitea_submit_pr_review,
+)
+
+FAKE_AUTH = "Basic dGVzdDp0ZXN0"
+SHA_PATTERN = re.compile(r"^llm-[0-9a-f]{12}$")
+# Two distinct, well-formed agent SHAs "belonging" to the same Gitea user.
+SHA_IMPLEMENTER = "llm-8f3a9c2d6b41"
+SHA_WOULD_BE_REVIEWER = "llm-41d0e7aa9f2c"
+
+
+def _pr(author, state="open", sha="abc123", mergeable=True):
+    return {
+        "user": {"login": author},
+        "state": state,
+        "head": {"sha": sha},
+        "mergeable": mergeable,
+    }
+
+
+class TestShaFormatConvention(unittest.TestCase):
+    """The documented format: llm-<12 lowercase hex>, nothing identifying."""
+
+    def test_documented_examples_are_valid(self):
+        for value in (SHA_IMPLEMENTER, SHA_WOULD_BE_REVIEWER, "llm-b7c93d441a08"):
+            self.assertRegex(value, SHA_PATTERN)
+
+    def test_identifying_or_malformed_values_are_rejected(self):
+        for bad in (
+            "llm-8F3A9C2D6B41",          # uppercase
+            "llm-8f3a9c2d6b4",           # too short
+            "llm-8f3a9c2d6b411",         # too long
+            "llm-opus4",                 # model name, not hex
+            "claude-8f3a9c2d6b41",       # provider prefix
+            "jcwalker3",                 # username
+            "llm-user@example.com",      # email
+            "8f3a9c2d6b41",              # missing prefix
+            "",
+        ):
+            self.assertNotRegex(bad, SHA_PATTERN)
+
+
+class TestShaCannotBypassSelfReview(unittest.TestCase):
+    """Scenario: user A (SHA 1) authored the PR; user A (SHA 2) tries to act."""
+
+    def _env(self, agent_sha, role):
+        # Reviewer-capable profile + a simulated launcher-injected agent SHA.
+        return {
+            "GITEA_PROFILE_NAME": "gitea-reviewer",
+            "GITEA_ALLOWED_OPERATIONS": "read,review,approve,merge",
+            "LLM_AGENT_SHA": agent_sha,
+            "LLM_AGENT_ROLE": role,
+        }
+
+    @patch("mcp_server.api_request")
+    @patch("mcp_server.get_auth_header", return_value=FAKE_AUTH)
+    def test_same_user_different_sha_cannot_approve(self, _auth, mock_api):
+        mock_api.side_effect = [{"login": "jcwalker3"}, _pr("jcwalker3")]
+        env = self._env(SHA_WOULD_BE_REVIEWER, "reviewer")
+        with patch.dict(os.environ, env, clear=True):
+            r = gitea_check_pr_eligibility(pr_number=9, action="approve", remote="prgs")
+        self.assertFalse(r["eligible"])
+        self.assertIn("authenticated user is PR author", r["reasons"])
+
+    @patch("mcp_server.api_request")
+    @patch("mcp_server.get_auth_header", return_value=FAKE_AUTH)
+    def test_same_user_different_sha_cannot_merge(self, _auth, mock_api):
+        mock_api.side_effect = [{"login": "jcwalker3"}, _pr("jcwalker3")]
+        env = self._env(SHA_WOULD_BE_REVIEWER, "merger")
+        with patch.dict(os.environ, env, clear=True):
+            r = gitea_check_pr_eligibility(pr_number=9, action="merge", remote="prgs")
+        self.assertFalse(r["eligible"])
+        self.assertIn("authenticated user is PR author", r["reasons"])
+
+    @patch("mcp_server.api_request")
+    @patch("mcp_server.get_auth_header", return_value=FAKE_AUTH)
+    def test_gated_merge_tool_refuses_self_merge_despite_sha(self, _auth, mock_api):
+        # Even the fully-confirmed gated merge path must refuse: correct
+        # confirmation string, mergeable PR, merge-capable profile — but the
+        # authenticated user is the PR author, whatever the agent SHA says.
+        mock_api.side_effect = [{"login": "jcwalker3"}, _pr("jcwalker3")]
+        env = self._env(SHA_WOULD_BE_REVIEWER, "merger")
+        with patch.dict(os.environ, env, clear=True):
+            r = gitea_merge_pr(
+                pr_number=9, confirmation="MERGE PR 9", remote="prgs")
+        self.assertFalse(r.get("performed"))
+        for call in mock_api.call_args_list:
+            method, url = call.args[0], call.args[1]
+            self.assertFalse(
+                method == "POST" and url.endswith("/merge"),
+                f"self-merge mutation reached the API: {method} {url}",
+            )
+
+    @patch("mcp_server.api_request")
+    @patch("mcp_server.get_auth_header", return_value=FAKE_AUTH)
+    def test_review_tool_refuses_self_approval_despite_sha(self, _auth, mock_api):
+        mock_api.side_effect = [{"login": "jcwalker3"}, _pr("jcwalker3")]
+        env = self._env(SHA_WOULD_BE_REVIEWER, "reviewer")
+        with patch.dict(os.environ, env, clear=True):
+            r = gitea_review_pr(
+                pr_number=9, event="APPROVE", body="self approve", merge=False,
+                remote="prgs")
+        self.assertFalse(r["success"])
+        self.assertIn("authenticated user is PR author", r["message"])
+        for call in mock_api.call_args_list:
+            method, url = call.args[0], call.args[1]
+            self.assertFalse(
+                method == "POST" and url.endswith("/reviews"),
+                f"self-review mutation reached the API: {method} {url}",
+            )
+
+
+class TestEligibilityNeverConsultsSha(unittest.TestCase):
+    """The gates have no SHA input: not via env, not via parameters."""
+
+    def _run_eligibility(self, extra_env):
+        env = {
+            "GITEA_PROFILE_NAME": "gitea-reviewer",
+            "GITEA_ALLOWED_OPERATIONS": "read,review,approve",
+        }
+        env.update(extra_env)
+        with patch("mcp_server.get_auth_header", return_value=FAKE_AUTH), \
+                patch("mcp_server.api_request") as mock_api:
+            mock_api.side_effect = [{"login": "reviewer-bot"}, _pr("author-bot")]
+            with patch.dict(os.environ, env, clear=True):
+                return gitea_check_pr_eligibility(
+                    pr_number=5, action="approve", remote="prgs")
+
+    def test_result_identical_with_without_and_across_shas(self):
+        baseline = self._run_eligibility({})
+        with_one = self._run_eligibility(
+            {"LLM_AGENT_SHA": SHA_IMPLEMENTER, "LLM_AGENT_ROLE": "implementer"})
+        with_other = self._run_eligibility(
+            {"LLM_AGENT_SHA": SHA_WOULD_BE_REVIEWER, "LLM_AGENT_ROLE": "reviewer"})
+        self.assertEqual(baseline, with_one)
+        self.assertEqual(baseline, with_other)
+        self.assertTrue(baseline["eligible"])  # sanity: a real decision ran
+
+    def test_eligibility_result_carries_no_agent_sha(self):
+        r = self._run_eligibility(
+            {"LLM_AGENT_SHA": SHA_IMPLEMENTER, "LLM_AGENT_ROLE": "implementer"})
+        blob = repr(r)
+        self.assertNotIn(SHA_IMPLEMENTER, blob)
+        self.assertNotIn("LLM_AGENT", blob)
+
+    def test_gate_functions_accept_no_agent_sha_parameter(self):
+        for fn in (gitea_check_pr_eligibility, gitea_merge_pr,
+                   gitea_review_pr, gitea_submit_pr_review):
+            for param in inspect.signature(fn).parameters:
+                lowered = param.lower()
+                self.assertNotIn("agent", lowered,
+                                 f"{fn.__name__} accepts agent param {param!r}")
+                self.assertNotIn("llm", lowered,
+                                 f"{fn.__name__} accepts llm param {param!r}")
+
+    def test_gate_sources_never_read_agent_sha(self):
+        # Phase 0 guarantee: no gate reads LLM_AGENT_* metadata at all.
+        for fn in (gitea_check_pr_eligibility, gitea_merge_pr,
+                   gitea_review_pr, gitea_submit_pr_review):
+            src = inspect.getsource(fn)
+            self.assertNotIn("LLM_AGENT", src,
+                             f"{fn.__name__} reads LLM_AGENT metadata")
+            self.assertNotIn("agent_sha", src,
+                             f"{fn.__name__} consults an agent SHA")
+
+
+if __name__ == "__main__":
+    unittest.main()