| name | dogfood-qa |
| description | QA the honeycomb CLI against a live account to find bugs and feature gaps. Target one command or resource family for manual testing, or sweep everything before a release. Use when dogfooding the CLI, reproducing a bug, or producing a findings report mapped to issues. |
| user-invocable | true |
Dogfood QA
Exercise honeycomb CLI commands against a live account to find bugs and feature gaps. Treat the account as disposable: mutate only throwaway resources.
Scope the Run
State the target before testing: one command (trigger update), one resource family (board plus board view), or a full sweep. Test only what is in scope. A sweep is this same method fanned out across every family; see references/sweep.md.
Test Matrix
For each command in scope:
--help matches actual behavior
- Happy-path CRUD
--format json and --format table
--no-interactive
- Error paths: missing required flags, bad IDs, invalid enums, malformed JSON (
-f -), nonexistent dataset
- Exit codes
- Flag names consistent with sibling commands
- Pagination
Trace each finding to cmd/<resource>/ or internal/ and cite file:line. File cosmetics too.
Capture Output
Do not judge exit codes or JSON through a pipe. cmd | head reports the pipe's exit status, not the command's, and 2>&1 | tee merges stderr into stdout. Both fabricate "exits 0 on error" and "corrupt JSON" findings. Capture the streams separately and read $? directly:
honeycomb <cmd> >out 2>err; echo "exit=$?"
jq -e . <out >/dev/null && echo valid || echo invalid
Safety
- Mutate only resources you create. Never touch existing account data.
- Run
create/update/delete against a throwaway environment and dataset.
- Use a throwaway profile for
auth login/logout.
references/setup.md covers the sandbox bypass, throwaway-environment provisioning, and teardown.
Record Findings
Write one markdown file per finding from assets/finding-template.md. For a sweep, group findings that share a root cause and build the report; see references/sweep.md.