dogfood-feature-defaults
Skill2 starsChanged 21 days agoMITView on GitHub
Decide default-on vs opt-in and breaking-change handling for new AgentLink features, settings, and behavior changes while AgentLink is still single-user dogfood. Use when adding a feature flag, experimental setting, new default behavior, settings or storage format change, or when asking whether to ship something off by default.
In this repository6 other files it gives its agents.
- AGENTS.mdAGENTS.md
- add-codex-modelSkill
- agentlink-dogfood-installSkill
- meridian-integrationSkill
- surface-paritySkill
- triageSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Default new behavior on. The point is to dogfood it.
- Distinctive: line 16
Prefer a design where any doubt falls back to the existing behavior, so turning it off is rarely needed.
- Distinctive: line 18
Do not ask Tristan whether something should be on by default.
- Very specific: line 25
Add telemetry that shows whether it helps, ideally a before/after signal visible in an existing npm run telemetry:* report.
- Very specific: line 26
Document it in the owning guide under resources/builtin-skills/documentation/references/, run npm run docs:generate, and add a CHANGELOG …
- Distinctive: line 27
Add the setting to the list below.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Dogfood Feature Defaults
AgentLink currently has no public usage: Tristan is the only user, building and testing it daily. Until Tristan explicitly says there is public usage, ship new behavior on by default with a way to turn it off, and do not spend effort protecting other users from breaking changes.
This policy is temporary. When Tristan says AgentLink has public users or that breaking changes now need care, stop applying it, tell him which experimental defaults are still on (see the list below), and ask how each should roll out.
While this policy applies
- Default new behavior on. The point is to dogfood it; an opt-in feature mostly goes untested.
- Give every default-on behavior that can misbehave an off switch, normally a boolean
agentlink.*setting. Start its description withExperimental:and say what turning it off restores. - Prefer a design where any doubt falls back to the existing behavior, so turning it off is rarely needed.
- Breaking changes to settings, stored session or telemetry formats, and internal contracts are acceptable. Do not add migrations, compatibility shims, or deprecation periods unless Tristan asks. Still keep reading old local data without crashing when that is cheap, such as tolerating a missing field.
- Do not ask Tristan whether something should be on by default. Default it on, and point out the off switch in the final summary.
- Everything else stays the same: validation gates, docs, CHANGELOG, and the repository's commit and push convention.
Checklist for a default-on experimental setting
package.json: add the setting with"default": true, the rightscope, and anExperimental:description.- Plumb it like
autoCondense:AgentConfiginsrc/agent/types.ts, theAgentSessionfield defaulting totrue, startup config and theonDidChangeConfigurationrefresh insrc/extension.ts, the resource-scoped snapshot insrc/agent/AgentSessionManagerHost.ts, andPROJECT_SCOPED_AGENTLINK_SETTINGSinsrc/adapters/vscode/projectSettingsAccessor.tsfor resource-scoped settings. Sessions copy config when they are created, so make the off switch reach open sessions too (seebackgroundCondenseinAgentSessionManager.updateConfig). - Add telemetry that shows whether it helps, ideally a before/after signal visible in an existing
npm run telemetry:*report. - Document it in the owning guide under
resources/builtin-skills/documentation/references/, runnpm run docs:generate, and add a CHANGELOG entry that names the setting for turning it off. - Add the setting to the list below.
Experimental defaults currently on
agentlink.backgroundCondense: prepares the auto-condense summary in the background before the threshold (src/agent/condensePreparation.ts). Measure it withnpm run telemetry:context(Condense timing section).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

