docs: LLM-Agent-SHA opaque attribution convention, Phase 0 (#86)

Implements the Phase 0 owner decision on #86 (issuecomment-1354): - docs/llm-agent-sha.md: format llm-<12 lowercase hex> (^llm-[0-9a-f]{12}$), generation rules, per-PR/workstream lifetime, visible markdown metadata blocks, no SHA in branch/worktree names, same-SHA vs same-user vs same-profile distinction. Attribution only — never an eligibility input. - docs/llm-workflow-runbooks.md: attribution subsection + handoff/review runbook pointers. - templates start-issue.md / review-pr.md: handoff and review metadata blocks; reviewer rule that a different SHA is not a different actor. - tests/test_llm_agent_sha.py: negative tests — same Gitea user with a different LLM-Agent-SHA still fails self-review and self-merge; eligibility results are identical with/without/across SHA env values; no gate accepts or reads any agent-SHA input. No launcher/env handling, no gitea_whoami fields, no PR auto-injection, no audit schema changes. No eligibility behavior changed. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-02 14:33:02 -04:00
parent 53e061bafd
commit 86141bfa0f
5 changed files with 374 additions and 2 deletions
@@ -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
@@ -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.
@@ -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.
 ```
@@ -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.
@@ -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()