Review: Forgejo MCP: add update_issue tool

Verdict: NEEDS_REFINEMENT

Template Completeness

• [x] Type — Feature
• [x] Lineage — standalone, discovered during dogfooding
• [x] Repo — forgejo_admin/claude-custom (BUT WRONG — see Repo Placement)
• [x] User Story — As Betty Sue, I want to update issue body via MCP
• [x] Context — explains consolidated spec convention gap
• [x] File Targets — two files listed (BUT WRONG paths — see File Targets)
• [x] Acceptance Criteria — 3 criteria
• [x] Test Expectations — unit + integration + run command
• [x] Constraints — API endpoint, patterns, optional fields
• [x] Checklist — standard PR/tests/no-unrelated
• [x] Related — references #161 and dotfiles#1

Traceability

• [x] story:pm-scope label — Betty Sue PM workflow, appropriate for MCP tooling
• [x] arch:forgejo-mcp label — correct architecture component
• [x] Forgejo issue — forgejo_admin/claude-custom#162, open — BUT filed on wrong repo (see Repo Placement)

File Targets

• [ ] ~/pal-e-docs-mcp/src/pal_e_docs_mcp/forgejo/tools.py — ISSUE: directory does not exist. ~/pal-e-docs-mcp is not present on the filesystem at all. The Forgejo MCP lives at ~/forgejo-mcp/src/forgejo_mcp/tools/workflows.py
• [ ] ~/pal-e-docs-mcp/src/pal_e_docs_mcp/forgejo/client.py — ISSUE: does not exist. The forgejo-mcp has no client.py — it uses forgejo_sdk.ForgejoClient via ~/forgejo-mcp/src/forgejo_mcp/server.py. No new client code is needed because the SDK already exposes issue_edit_issue(owner, repo, index, *, title=None, body=None, ...)

• ~/forgejo-mcp/src/forgejo_mcp/tools/workflows.py — add update_issue tool function following existing patterns (e.g., comment_on_issue)
• ~/forgejo-mcp/tests/test_label_comment_repo.py (or new test file) — add integration tests

Repo Placement

Dependencies

• forgejo-sdk issue_edit_issue — VERIFIED. The SDK method exists with signature: (owner, repo, index, *, title=None, body=None, assignee=None, assignees=None, due_date=None, milestone=None, ref=None, state=None, ...). Both title and body are optional kwargs. No SDK changes needed.
• Board item #364 (issue #161, "Scope review pipeline") — upstream motivation. Not a hard blocker.
• Board item #360 (forgejo-mcp#15) — DUPLICATE. Same scope, correct repo.

Acceptance Criteria

• [x] mcp__forgejo__update_issue(owner, repo, issue_number, body=..., title=...) — testable, clear
• [x] Can update body without title and vice versa — testable, SDK supports optional kwargs
• [x] Updated body appears on Forgejo web UI — integration test can verify via API round-trip

Blast Radius

• Low blast radius. This adds a new tool following established patterns. No existing tools are modified.
• The existing codebase already uses issue_edit_issue in test cleanup (closing issues), confirming the SDK method is stable.
• No downstream consumers need changes — MCP tool registration is automatic via the @mcp.tool() decorator pattern.

Decomposition (5-minute rule)

• 1 file target for implementation + 1 for tests, single repo — no decomposition needed
• 3 acceptance criteria — within threshold
• Estimated agent work: under 5 minutes — fits single agent pass

Recommendation

• Close claude-custom#162 as duplicate, linking to forgejo-mcp#15
• Remove board item #365 from the board (duplicate of #360)
• Update forgejo-mcp#15 with corrected file targets:
• Confirm forgejo-mcp#15 has the SDK signature detail: client.issue_edit_issue(owner, repo, index, title=..., body=...) — no client.py changes needed