Files
leo-claude-mktplace/plugins/projman/commands/debug-report.md
lmiranda b7829dca05 fix: add curl fallback to /debug-report when MCP tools unavailable
When MCP tools are not available in a session (the very scenario
/debug-report is designed to diagnose), the command now falls back to:

1. Check for Gitea credentials at ~/.config/claude/gitea.env
2. Use curl + jq to create the issue via Gitea REST API
3. If no credentials, save report to local file for manual submission

Security measures:
- Uses mktemp -m 600 for restrictive file permissions
- Uses jq --rawfile for safe JSON construction (no command substitution)
- Proper cleanup of temporary files

Fixes #100

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-22 14:19:05 -05:00

8.2 KiB

description
description
Run diagnostics and create structured issue in marketplace repository

Debug Report

Run diagnostic checks on projman MCP tools and create a structured issue in the marketplace repository for investigation.

Prerequisites

Your project .env must have:

PROJMAN_MARKETPLACE_REPO=personal-projects/leo-claude-mktplace

If not configured, ask the user for the marketplace repository path.

CRITICAL: Execution Steps

You MUST follow these steps in order. Do NOT skip any step.

Step 1: Gather Project Context

Run these Bash commands to capture project information:

# Get git remote URL
git remote get-url origin

# Get current branch
git branch --show-current

# Get working directory
pwd

Parse the git remote to extract REPO_NAME in owner/repo format.

Store all values:

  • PROJECT_REPO: The detected owner/repo
  • GIT_REMOTE: Full git remote URL
  • CURRENT_BRANCH: Current branch name
  • WORKING_DIR: Current working directory

Step 2: Read Marketplace Configuration

grep PROJMAN_MARKETPLACE_REPO .env

Store as MARKETPLACE_REPO. If not found, ask the user.

Step 3: Run Diagnostic Suite

Run each MCP tool with explicit repo parameter. Record success/failure and full response.

Test 1: validate_repo_org

mcp__plugin_projman_gitea__validate_repo_org(repo=PROJECT_REPO)

Expected: {is_organization: true/false}

Test 2: get_labels

mcp__plugin_projman_gitea__get_labels(repo=PROJECT_REPO)

Expected: {organization: [...], repository: [...], total_count: N}

Test 3: list_issues

mcp__plugin_projman_gitea__list_issues(repo=PROJECT_REPO, state="open")

Expected: Array of issues

Test 4: list_milestones

mcp__plugin_projman_gitea__list_milestones(repo=PROJECT_REPO)

Expected: Array of milestones

Test 5: suggest_labels

mcp__plugin_projman_gitea__suggest_labels(context="Test bug fix for authentication")

Expected: Array of label names matching repo's format

For each test, record:

  • Tool name
  • Exact parameters used
  • Status: PASS or FAIL
  • Response or error message

Step 4: Analyze Results

Count failures and categorize errors:

Category Indicators
Parameter Format "owner/repo format", "missing parameter"
Authentication "401", "403", "unauthorized"
Not Found "404", "not found"
Network "connection", "timeout", "ECONNREFUSED"
Logic Unexpected response format, wrong data

For each failure, write a hypothesis about the likely cause.

Step 5: Generate Issue Content

Use this exact template:

## Diagnostic Report

**Generated**: [ISO timestamp]
**Command Tested**: [What the user was trying to run, or "general diagnostic"]
**Reporter**: Claude Code via /debug-report

## Project Context

| Field | Value |
|-------|-------|
| Repository | `[PROJECT_REPO]` |
| Git Remote | `[GIT_REMOTE]` |
| Working Directory | `[WORKING_DIR]` |
| Current Branch | `[CURRENT_BRANCH]` |

## Diagnostic Results

### Test 1: validate_repo_org

**Call**: `validate_repo_org(repo="[PROJECT_REPO]")`
**Status**: [PASS/FAIL]
**Response**:
```json
[full response or error]

Test 2: get_labels

Call: get_labels(repo="[PROJECT_REPO]") Status: [PASS/FAIL] Response:

[full response or error - truncate if very long]

[... repeat for each test ...]

Summary

  • Total Tests: 5
  • Passed: [N]
  • Failed: [N]

Failed Tools

[List each failed tool with its error]

Error Category

[Check applicable categories]

Hypothesis

[Your analysis of what's wrong and why]

Suggested Investigation

  1. [First file/function to check]
  2. [Second file/function to check]
  3. [etc.]

Reproduction Steps

  1. Navigate to [WORKING_DIR]
  2. Run [command that was being tested]
  3. Observe error at step [X]

Generated by /debug-report - Labels: Type: Bug, Source: Diagnostic, Agent: Claude


### Step 6: Create Issue in Marketplace

**First, check if MCP tools are available.** Attempt to use an MCP tool. If you receive "tool not found", "not in function list", or similar error, the MCP server is not accessible in this session - use the curl fallback.

#### Option A: MCP Available (preferred)

mcp__plugin_projman_gitea__create_issue( repo=MARKETPLACE_REPO, title="[Diagnostic] [summary of main failure]", body=[generated content from Step 5], labels=["Type: Bug", "Source: Diagnostic", "Agent: Claude"] )


If labels don't exist, create issue without labels.

#### Option B: MCP Unavailable - Use curl Fallback

If MCP tools are not available (the very issue you may be diagnosing), use this fallback:

**1. Check for Gitea credentials:**

```bash
if [[ -f ~/.config/claude/gitea.env ]]; then
  source ~/.config/claude/gitea.env
  echo "Credentials found. API URL: $GITEA_API_URL"
else
  echo "No credentials at ~/.config/claude/gitea.env"
fi

2. If credentials exist, create issue via curl with proper JSON escaping:

Create secure temp files and save content:

# Create temp files with restrictive permissions
DIAG_TITLE=$(mktemp -p /tmp -m 600 diag-title.XXXXXX)
DIAG_BODY=$(mktemp -p /tmp -m 600 diag-body.XXXXXX)
DIAG_PAYLOAD=$(mktemp -p /tmp -m 600 diag-payload.XXXXXX)

# Save title
echo "[Diagnostic] [summary of main failure]" > "$DIAG_TITLE"

# Save body (paste Step 5 content) - heredoc delimiter prevents shell expansion
cat > "$DIAG_BODY" << 'DIAGNOSTIC_EOF'
[Paste the full issue content from Step 5 here]
DIAGNOSTIC_EOF

Construct JSON safely using jq's --rawfile (avoids command substitution):

# Build JSON payload using jq with --rawfile for safe content handling
jq -n \
  --rawfile title "$DIAG_TITLE" \
  --rawfile body "$DIAG_BODY" \
  '{title: ($title | rtrimstr("\n")), body: $body}' > "$DIAG_PAYLOAD"

# Create issue using the JSON file
curl -s -X POST "${GITEA_API_URL}/repos/${MARKETPLACE_REPO}/issues" \
  -H "Authorization: token ${GITEA_API_TOKEN}" \
  -H "Content-Type: application/json" \
  -d @"$DIAG_PAYLOAD" | jq '.html_url // .'

# Secure cleanup
rm -f "$DIAG_TITLE" "$DIAG_BODY" "$DIAG_PAYLOAD"

3. If no credentials found, save report locally:

REPORT_FILE=$(mktemp -p /tmp -m 600 diagnostic-report-XXXXXX.md)
cat > "$REPORT_FILE" << 'DIAGNOSTIC_EOF'
[Paste the full issue content from Step 5 here]
DIAGNOSTIC_EOF
echo "Report saved to: $REPORT_FILE"

Then inform the user:

MCP tools are unavailable and no Gitea credentials found at ~/.config/claude/gitea.env.

Diagnostic report saved to: [REPORT_FILE]

To create the issue manually:
1. Configure credentials: See docs/CONFIGURATION.md
2. Or create issue directly at: http://gitea.hotserv.cloud/[MARKETPLACE_REPO]/issues/new

Step 7: Report to User

Display summary:

Debug Report Complete
=====================

Project: [PROJECT_REPO]
Tests Run: 5
Passed: [N]
Failed: [N]

Failed Tools:
  - [tool1]: [brief error]
  - [tool2]: [brief error]

Issue Created: [issue URL]

Next Steps:
  1. Switch to marketplace repo: cd [marketplace path]
  2. Run: /debug-review
  3. Select issue #[N] to investigate

DO NOT

  • DO NOT attempt to fix anything - only report
  • DO NOT create issues if all tests pass (just report success)
  • DO NOT skip any diagnostic test
  • DO NOT call MCP tools without the repo parameter
  • DO NOT ask user questions during execution - run autonomously

If All Tests Pass

If all 5 tests pass, report success without creating an issue:

Debug Report Complete
=====================

Project: [PROJECT_REPO]
Tests Run: 5
Passed: 5
Failed: 0

All diagnostics passed. No issues to report.

If you're experiencing a specific problem, please describe it
and I can create a manual bug report.

Troubleshooting

PROJMAN_MARKETPLACE_REPO not configured

  • Ask user: "What is the marketplace repository? (e.g., personal-projects/leo-claude-mktplace)"
  • Store for this session and remind user to add to .env

Cannot detect project repository

  • Check if in a git repository: git rev-parse --git-dir
  • If not a git repo, ask user for the repository path

MCP tools not available

  • Use the curl fallback in Step 6, Option B
  • Requires Gitea credentials at ~/.config/claude/gitea.env
  • If no credentials, report will be saved locally for manual submission