chore: neutralize Claude-specific branding for multi-provider positioning

[dean0x]

Summary

• Replace Claude-specific language with provider-neutral terminology across all public-facing text
• Updates README, package.json (keywords + description), FEATURES.md, SETUP_GUIDE.md, bug report template, CONTRIBUTING.md, and MCP tool descriptions
• Internal class names (ClaudeProcessSpawner, claude-adapter.ts) and historical release notes intentionally unchanged

Motivation

Backbeat added multi-agent support in v0.5.0 (Claude, Codex, Gemini), but the branding still read as Claude-only. This positions Backbeat as provider-neutral ahead of OpenAI's open source program submission and broader multi-provider appeal.

Files changed (7)

File	Change
README.md	Problem statement, prerequisites, troubleshooting → agent-neutral
package.json	Keywords: claude/claude-code → ai-agents/multi-agent/codex/gemini; description updated
docs/FEATURES.md	"Claude Code Integration" → "Agent Process Management"; DelegateTask description neutralized
docs/SETUP_GUIDE.md	Title, session references, CLI section → neutral
.github/ISSUE_TEMPLATE/bug_report.md	"Claude Code console/version" → "MCP client/agent CLI"
CONTRIBUTING.md	Removed "from Claude Code" qualifier
src/adapters/mcp-adapter.ts	DelegateTask tool description + prompt field description → neutral

Test plan

• npm run build — clean (no compile errors from mcp-adapter change)
• Visual review of README.md diff for tone/flow
• CI passes

[greptile-apps]

Confidence Score: 4/5

• Safe to merge with minor documentation fixes — no runtime or logic changes are present.
• All source code changes are limited to string literals in MCP tool descriptions with no behavioral impact. The documentation issues (duplicate lines, missed branding) are purely cosmetic but the duplicate content in SETUP_GUIDE.md is user-visible and should be fixed before merging.
• docs/SETUP_GUIDE.md (duplicate lines at 101–107, hardcoded claude at line 54) and package.json (Claude-specific copy in test script at line 18).

Important Files Changed

Filename	Overview
docs/SETUP_GUIDE.md	Title and section headings neutralized, but the PR introduced duplicate lines (intro paragraph and step 3 both appear twice), and the Step 3 example command still hardcodes claude.
package.json	Keywords and description successfully neutralized, but the test script warning message and test:audit script still contain Claude-specific copy ("crashes Claude Code instances", "run these from Claude Code") that was missed.
README.md	Problem statement, prerequisites, MCP tools section, and troubleshooting all correctly updated to provider-neutral language; versioned roadmap entries and code examples unchanged as intended.
docs/FEATURES.md	"Claude Code Integration" section renamed to "Agent Process Management" with all bullet points correctly neutralized; historical release notes untouched as intended.
src/adapters/mcp-adapter.ts	DelegateTask tool description and prompt field description correctly updated to provider-neutral language; no logic or schema changes, purely cosmetic text update.
.github/ISSUE_TEMPLATE/bug_report.md	Log source and environment version field cleanly updated from Claude-specific to MCP-client/agent-agnostic wording.
CONTRIBUTING.md	Single comment line updated from "Safe to run from Claude Code or any environment" to "Safe to run individually" — clean and consistent with the neutralization goal.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[PR: Neutralize Claude Branding] --> B[Public-Facing Text Files]
    A --> C[Source Code]

    B --> D[README.md\nProblem statement, prerequisites,\ntroubleshooting]
    B --> E[docs/FEATURES.md\nSection headings + bullet points]
    B --> F[docs/SETUP_GUIDE.md\nTitle, step headings, CLI section]
    B --> G[package.json\nkeywords + description]
    B --> H[.github/ISSUE_TEMPLATE/bug_report.md\nLog sources + environment field]
    B --> I[CONTRIBUTING.md\nTest comment line]

    C --> J[src/adapters/mcp-adapter.ts\nDelegateTask + prompt descriptions]

    F --> F1["⚠️ Duplicate lines introduced\n(lines 101–102, 106–107)"]
    F --> F2["⚠️ Step 3 still hardcodes 'claude' CLI"]
    G --> G1["⚠️ test script warning still\nreferences 'Claude Code instances'"]

    J --> K["✅ No logic changes\nPure text update"]

    style F1 fill:#ffcccc
    style F2 fill:#ffcccc
    style G1 fill:#ffcccc
    style K fill:#ccffcc

Loading

Comments Outside Diff (2)

1. package.json, line 18 (link)

   Claude-specific copy missed in test script warning

   package.json is listed as a changed file in this PR, and its description and keywords were updated, but the test script warning still contains two hardcoded Claude references:

   • "crashes Claude Code instances" → should be something like "crashes AI coding agent instances"
   • "Safe commands (run these from Claude Code)" → should be "Safe commands (run these individually)" (the same wording used for the equivalent comment in CONTRIBUTING.md)

   The test:audit script on line 41 also still reads 'Run /audit-tests command in Claude for full audit', which is Claude-specific. If the goal is consistent provider-neutral language in all public-facing text, these three copy strings in package.json should be updated alongside the description and keywords changes already made.

2. docs/SETUP_GUIDE.md, line 54 (link)

   Step 3 heading neutralized but example command still hardcodes claude

   The section heading was changed from "Start NEW Claude Code Session" to "Start a New Coding Session", but the example command directly below it still reads claude "Let's test Backbeat MCP tools". For a provider-neutral guide this is inconsistent — a Codex or Gemini user following Step 3 would see a Claude-specific command even after the neutral heading.

   Consider adding a note or showing the provider-specific variant, e.g.:

   Or replace the example with a generic instruction: "Start your agent CLI with a new session and test the MCP tools."

_{Last reviewed commit: d049b42}

[greptile-apps]

Duplicate lines introduced by this PR

The edit accidentally left both the old and new text. The paragraph intro appears twice (lines 101–102) and "3. Test with real tasks" is listed twice as consecutive step items (lines 106–107). The section should read:

Suggested change

	To use with an actual agent CLI (not mock):
	To use with an actual agent CLI (not mock):
	1. Remove `"MOCK_MODE": "true"` from config
	2. Ensure `claude` CLI is installed
	2. Ensure your chosen agent CLI is installed (e.g., `claude`, `codex`, `gemini-cli`)
	3. Test with real tasks
	3. Test with real tasks
	To use with an actual agent CLI (not mock):
	1. Remove `"MOCK_MODE": "true"` from config
	2. Ensure your chosen agent CLI is installed (e.g., `claude`, `codex`, `gemini-cli`)
	3. Test with real tasks