OpenParable
Skip to content
← Back to skills

Reproduce Fuzz Crash

Testing & Quality
reproduce-fuzz-crash

Reproduce Fuzz Crash is an Agent Skill for testing & quality published in the dragonflydb/dragonfly repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Reproduce AFL++ fuzz crashes from GitHub Actions. Use when user provides a GitHub Actions fuzz run URL and wants to reproduce and analyze the crash locally.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/dragonflydb/dragonfly/blob/HEAD/.claude/skills/reproduce-fuzz-crash/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/reproduce-fuzz-crash, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Reproduce Fuzz Crash

Given a GitHub Actions fuzz run URL, download crash artifacts, triage them with fuzz/triage_crashes.sh, and produce a crash analysis report.

Input: $ARGUMENTS — a GitHub Actions run URL like: https://github.com/dragonflydb/dragonfly/actions/runs/22906484769 or with query params like ?pr=6855.

Workflow

Step 1: Parse the URL

Extract owner/repo and run_id from the URL.

https://github.com/{owner}/{repo}/actions/runs/{run_id}[?...]

Strip any query parameters from run_id.

Step 2: Download artifacts

List crash artifacts via the GitHub API, then download each as a .zip directly:

IMPORTANT: Run each command as a separate Bash tool call (no && chaining) to ensure auto-approval works with the user's permission patterns.

# List artifacts — filter for names containing "crash"
gh api repos/{owner}/{repo}/actions/runs/{run_id}/artifacts

# Create output directory
mkdir -p /tmp/fuzz-repro-{run_id}

# Download each crash artifact by ID (separate command)
gh api repos/{owner}/{repo}/actions/artifacts/{artifact_id}/zip > /tmp/fuzz-repro-{run_id}/<artifact-name>.zip

This gives real .zip files that the triage script can consume directly.

If no crash artifacts are found, report that the run has no crash artifacts and stop.

Note: there may be duplicate artifact names (same name, different IDs) from retried jobs. Download the most recent one (highest artifact ID).

Step 3: Determine mode

Infer the protocol mode from the artifact name:

  • Contains "memcache" → memcache
  • Otherwise → resp

Step 4: Check Dragonfly binary

Check if the debug binary already exists and runs:

./build-dbg/dragonfly --version

Only build if the binary doesn't exist or fails to run:

ninja -C build-dbg dragonfly

If build-dbg doesn't exist, run ./helio/blaze.sh first.

Step 5: Run triage_crashes.sh

For each zip file, run:

./fuzz/triage_crashes.sh ./build-dbg/dragonfly <mode> /tmp/fuzz-repro-{run_id}/<artifact-name>.zip

Capture the full output.

Step 6: Analyze and report

Parse the triage output for confirmed crashes. For each confirmed crash:

  1. Read the source at the crash location — use the stack trace to identify the source file and line number, then read that code.
  2. Provide analysis: likely root cause, what to investigate.

Print a structured report:

## Fuzz Crash Report

**Run**: {url}
**Artifacts**: {number} crash(es) found

---

### Crash NNNNNN

**Reproduced**: Yes / No (false positive)
**Signal**: SIGABRT (6) / SIGSEGV (11) / etc.

**Stack trace**:
\```
<stack trace from triage output>
\```

**Analysis**:
<1-3 sentences explaining the likely root cause based on the stack trace,
the assertion message, and the crash input. Identify the source file and
line number. Suggest what to investigate.>

Important Notes

  • The triage script uses port 6379 (resp) or 11211 (memcache). Ensure no other Dragonfly or Redis instance is using these ports.
  • The script adds --restricted_commands flags to avoid false positives from commands like DEBUG SLEEP that the fuzzer might generate.
  • Some crashes are non-deterministic (thread timing). The script reports these as "FALSE POSITIVE" — note this clearly, it doesn't mean the bug is invalid, just that it didn't reproduce on this run.
  • The script handles its own cleanup of Dragonfly processes.
  • Do NOT delete /tmp/fuzz-repro-{run_id}/ — the user may want to inspect it.
  • If gh run download fails with permissions, suggest the user authenticate with gh auth login.