chore(ai): add agent notes, Cursor rules, and skills - #2133
Conversation
Give coding agents a local map of commands, invariants, and review skills so they follow project conventions without copying onboarding. Signed-off-by: Felipe Bergamin <felipebergamin@profusion.mobi>
|
@alanpeixinho please feel free to suggest your own rules and skills that you find relevant |
|
#1719 also brings some skills on it, might be worth it reviewing. |
| @@ -0,0 +1,40 @@ | |||
| #!/usr/bin/env bash | |||
There was a problem hiding this comment.
what exatcly is this sript?
There was a problem hiding this comment.
it is part of the skill. The agent can use it to append logs by invoking the script
|
|
||
| ## Git | ||
|
|
||
| Conventional Commits with a scope, e.g. `fix(tree-compare): restore drawer navigation`. One concern per commit/PR. PR titles match; describe why, how, trade-offs, tests; link issues; screenshots/GIFs for UI. CI green before asking for review. Commit with "Signed-off-by". |
There was a problem hiding this comment.
we could include reminder to a github closes tags
There was a problem hiding this comment.
good point. I will update with this
| 4. Pass each subagent the same scoped diff/file context and ask it to return prioritized findings with file references and evidence. | ||
| 5. After both finish, synthesize the results with findings first, deduplicated across reviewers. Weight overlapping findings more heavily, resolve disagreements with your own judgment, and keep summaries brief. | ||
|
|
||
| If individual background summaries are already visible to the user, do not restate them wholesale. Surface the unified verdict, the highest-signal findings, and any remaining uncertainty. |
There was a problem hiding this comment.
maybe worth including a guard to ask for human intervention for review after it finishes running pre-review
| @@ -0,0 +1,16 @@ | |||
| --- | |||
There was a problem hiding this comment.
might be worth keeping a kci-dashboard folder, so we can know which skills are specific written for the project
| @@ -0,0 +1,68 @@ | |||
| --- | |||
There was a problem hiding this comment.
is it worth keeping a non development related skill?
There was a problem hiding this comment.
I think so. This skill is referenced by another ones (e.g. show-me-your-work) and we can also use to generate summaries, issues, ai reviews, etc.
alanpeixinho
left a comment
alanpeixinho
left a comment
There was a problem hiding this comment.
Left some comments, none are blockers
| disable-model-invocation: true | ||
| --- | ||
|
|
||
| # Unslop |
There was a problem hiding this comment.
|
Did you generate all of this yourself or did you copy it from somewhere? I'm concerned you might have forgotten proper attribution |
|
I might've missed, but I'd also like "Assisted-by:" tags in commits heavily influenced by LLMs, to alert humans reviewing the code that it could contain hallucinations, or generally inform future readers about it |
skills are from agents.md is mine. |
d3bfa33 to
68fe12c
Compare
rules are also mine. except for ponytail |
Give agents a short index of vendored skills and Cursor rules before the rest of the notes. Signed-off-by: Felipe Bergamin <felipebergamin@profusion.mobi>
3 participants
Summary
AGENTS.mdwith where-to-look, commands, when/how to run the app, Postgres caveats, and project invariants for coding agents..agentsskills andskills-lock.json.Human onboarding stays in
docs/Onboarding.md/CONTRIBUTING.md; this is the short agent-facing map.Test plan
AGENTS.mdand confirm commands/URLs matchdocs/dev-environment.md(make dev, ports 9000/5173/8000)..cursor/rules/and.agents/skills/load in Cursor for this repo.