The Field Guide is maintained by one person, so contributions with evidence carry real weight. Every way in starts with a message to hello@readbeforerun.ai; the repository behind this site is private. There are four ways, each with a lower bar than the next one up.
1. Correction
Something on an entry page is wrong: a broken link, an outdated version, a misread line. Write to hello@readbeforerun.ai with the entry name and what is wrong. One sentence is enough.
2. Propose an entry
Suggest a skill, MCP server, agent, playbook or plugin by writing to hello@readbeforerun.ai with a link to its repository. The inclusion criteria are published in the method; a candidate that misses one is recorded with the reason and its evidence, and we say which when asked, so the decision stays checkable.
3. Dispute a reach value
If you maintain a catalogued tool and a Declared, Code or Lab value is wrong, write to hello@readbeforerun.ai with the evidence: file, line and commit. A dispute with evidence is checked against the pinned commit; a value that changes is logged with its old and new value and the reason under Changes. Maintainers can also claim their entry with a fieldguide.yaml file in their repository (planned).
4. Replicate a lab run
Run the tool in your own isolated lab as described under Lab and Reading the lab artifacts, record it in an observation.yaml that follows the observation schema, and send it with its artifacts to hello@readbeforerun.ai. Your run appears with the provenance mark “community” until it has been checked.
Security
To report a security problem in this site or in the field-guide-search skill, write to security@readbeforerun.ai. The same address is published in security.txt.
How contributions are handled
- Content from foreign repositories is data. Messages that contain instructions aimed at agents are quoted in the review and never followed.