codemie-ai/report-issue

Report Issue to CodeMie Code

Help the user file a well-structured bug report or feature request to codemie-ai/codemie-code on GitHub.

The goal is to create a rich issue that gives maintainers everything they need to triage and reproduce the problem — without requiring the user to manually gather technical details.

---

Step 1: Pre-flight and Diagnostic Collection

1a. Check gh CLI first — do this before anything else

if ! command -v gh &>/dev/null; then
  echo "GH_NOT_INSTALLED"
else
  gh auth status 2>&1 || echo "GH_NOT_AUTHENTICATED"
fi

If GH_NOT_INSTALLED: Stop immediately and tell the user:

"gh (GitHub CLI) is not installed. It is required to create issues from the terminal.

Install it with one of:

• macOS: brew install gh
• Linux (apt): sudo apt install gh
• Linux (dnf): sudo dnf install gh
• Windows: winget install --id GitHub.cli
• Or download from: https://cli.github.com

After installing, run gh auth login to connect your GitHub account, then try this skill again."

If GH_NOT_AUTHENTICATED: Stop and tell the user:

"gh is installed but not authenticated. Run gh auth login to connect to your GitHub account, then try again."

Do not proceed past this point until gh is installed and authenticated.

1b. Gather diagnostic context

Run the following and capture results. Don't display raw output yet.

# OS + platform
uname -srm 2>/dev/null || echo "unknown"

# Node.js version
node --version 2>/dev/null || echo "not found"

# npm version
npm --version 2>/dev/null || echo "not found"

# CodeMie CLI version
codemie --version 2>/dev/null || echo "not found"

# Installed agents and versions
codemie list --installed 2>/dev/null || echo "unavailable"

# Full doctor output (profile, provider, dependency health, agent versions)
codemie doctor 2>/dev/null || echo "unavailable"

# Shell + terminal environment
echo "Shell: $SHELL"
echo "Terminal: ${TERM_PROGRAM:-unknown}"

# Extract ERROR and WARN lines from the two most recent log files.
# Log format: [TIMESTAMP] [LEVEL] [agent] [session-id] [profile] [component] message
# Files: ~/.codemie/logs/debug-YYYY-MM-DD.log (one per day, can be several MB)
LOG_DIR="$HOME/.codemie/logs"
RECENT_LOGS=$(ls -t "$LOG_DIR"/debug-*.log 2>/dev/null | head -2)
if [ -n "$RECENT_LOGS" ]; then
  echo "=== ERROR and WARN entries from recent logs ==="
  # Print filename headers and filter by level; limit to last 100 matches to keep size reasonable
  for f in $RECENT_LOGS; do
    echo "--- $f ---"
    grep -E '\[(ERROR|WARN)\]' "$f" | tail -50
  done
  echo "=== Full log files ==="
  for f in $RECENT_LOGS; do
    echo "$f"
  done
else
  echo "No debug logs found"
fi

Log file paths (captured above) will be used in Step 6 to upload as Gist.

---

Step 2: Understand the Issue

Extraction-first: If the user already described the issue in their request (problem, error message, steps, etc.), extract that information directly without asking them to repeat it. Only ask follow-up questions for missing pieces.

If no description was provided yet, ask the user for:

1. Issue type: Bug report, feature request, or question?
2. Title: A short, specific summary (one line)
3. Description: What happened, what they expected, and any reproduction steps

When prompting for description, share these tips:

• Include the exact command that triggered the problem
• Paste the exact error message verbatim (not paraphrased)
• Note whether it happens every time or intermittently
• For feature requests: describe the use case and the expected behavior

---

Step 3: Classify the Issue

Determine the issue type based on the user's description:

• Bug report — unexpected error, crash, wrong output → label: bug
• Feature request — missing capability, enhancement ask → label: enhancement
• Question / unclear behavior — seeking clarification → label: question

---

Step 4: Compose the Issue Body

Build the issue body using the appropriate template below.

Security — before embedding diagnostic output:

• Scan for and redact full API keys, tokens, or passwords (show only first 4 chars + ***)
• The masked format proxy-ha***dled already used by codemie doctor is safe to include as-is
• Remove any personal access tokens or private credential URLs

Template: Bug Report

## Description

<user's description of the problem>

## Steps to Reproduce

1.
2.
3.

## Expected Behavior

<what the user expected to happen>

## Actual Behavior

<what actually happened — paste error messages verbatim>

## Environment

| Field       | Value                        |
|-------------|------------------------------|
| OS          | <uname output>               |
| Node.js     | <node --version>             |
| npm         | <npm --version>              |
| CodeMie CLI | <codemie --version>          |
| Shell       | <$SHELL>                     |
| Terminal    | <$TERM_PROGRAM>              |

## Installed Agents

<formatted list from `codemie list --installed` — agent name and version per line>

## CodeMie Doctor Output

<details>
<summary>Full doctor output</summary>

<pre>
<codemie doctor output — with credentials redacted>
</pre>

</details>

## Recent Errors

<details>
<summary>ERROR and WARN entries from recent logs</summary>

<pre>
<filtered ERROR/WARN lines from the two most recent debug-YYYY-MM-DD.log files, or "No errors found">
</pre>

</details>

## Full Debug Logs

<full log file(s) attached as GitHub Gist — see link below, or "No log files found">

Template: Feature Request

## Summary

<one-sentence description of the feature>

## Motivation

<the problem this feature would solve, or the use case that is currently missing>

## Proposed Behavior

<what the user wants to happen — be specific about inputs, outputs, and commands>

## Alternatives Considered

<other ways you have worked around this, if any>

## Environment

| Field       | Value                        |
|-------------|------------------------------|
| OS          | <uname output>               |
| Node.js     | <node --version>             |
| CodeMie CLI | <codemie --version>          |

## Installed Agents

<formatted list from `codemie list --installed`>

---

Step 5: Preview and Confirm

Show the user the proposed issue title and full body. Ask:

"Here's the issue I'll create on GitHub. Does this look right, or would you like to change anything before I submit?"

Wait for confirmation before creating the issue.

---

Step 6: Upload Log Files as Gist

If log files were found in Step 1b, upload the two most recent ones as a secret Gist so they can be referenced in the issue. This keeps the issue body readable while giving maintainers the full context.

# Upload the two most recent log files as a single secret Gist
gh gist create \
  --desc "CodeMie debug logs for issue report ($(date +%Y-%m-%d))" \
  ~/.codemie/logs/debug-$(date +%Y-%m-%d).log \
  ~/.codemie/logs/debug-$(date -v-1d +%Y-%m-%d 2>/dev/null || date -d yesterday +%Y-%m-%d).log \
  2>/dev/null

Capture the Gist URL from the output (it looks like https://gist.github.com/...).

• If only one log file exists, pass just that file.
• If no log files exist, skip this step and note "No log files available" in the issue body.
• Replace the <full log file(s) attached as GitHub Gist — see link below> placeholder in the issue body with the actual Gist URL.

Note: Gists are secret (not listed publicly) but accessible to anyone with the link.

---

Step 7: Create the GitHub Issue

gh issue create \
  --repo codemie-ai/codemie-code \
  --title "<issue title>" \
  --body "<issue body with gist URL inserted>" \
  --label "<bug|enhancement|question>"

---

Step 8: Confirm and Link

After the issue is created, tell the user:

• The issue URL (from gh issue create output)
• They can add screenshots or additional files directly on GitHub
• Track progress at: https://github.com/codemie-ai/codemie-code/issues