Sabitlenmiş Tweet
plannotator
580 posts
plannotator
@plannotator
Annotate agent plans and review code visually. Share with your team, iterate, and send feedback to your agent with one click • https://t.co/3Gg5SohuqE
Katılım Ocak 2026
39 Takip Edilen1.1K Takipçiler
Plannotator 0.19.17
- setup goal skill reworked to simpler facts-based
- added `plannotator --version`
English
plannotator retweetledi
The more i use `/goal` the more I want to ensure the model gets simple facts right up front. So I use a version of @mattpocockuk grill-me focused on @Everlier's methodology around facts.
Facts allow me to describe very clear requirements for a feature or system. & Not always that technical. The more I do with AI, the less patient I get, so I can't read a ton of markdown, but I can afford to read/verify simple lists of facts.
Once facts are established i let the agent refine a plan for itself (every decision aligned to those facts) so it can figure out order of operations.
have been having a lot of fun with this process.
composition by @editframe
English
@shinedairz Ive been iterating a bunch on this & have refined the skill to a much simpler, yet maybe more powerful state. A lot easier for you to review - focuses on facts of the thing you want.
github.com/backnotprop/pl…
English
@plannotator Hey, just wanted to say that I'm having a blast using this more and more. To give some fb I noticed then that whenever I use the goal setup skill within codex, it will sometimes be impatient and run "plannotator sessions --open 1" while I'm still reading - can this be mitigated?
English
plannotator retweetledi
built a skill that reorders any diff into a narrative walkthrough, then opens it in @plannotator so you can annotate inline. The agent reads your comments and fixes the code, then regenerates the walkthrough so you can go again.
Point it at a remote PR and it'll still generate the walkthrough (just without the fix loop)
English
@pierrecomputer cmd/ctrl+click on tokens to find where they are defined or used
English
Now in Plannotator code review - code refs
maybe the spiciest diffs in all of the land (diffs lib by @pierrecomputer)
English
@pierrecomputer *npm had an outage earlier... now released for Pi and OpenCode.
English
@pyrons_ @JinjingLiang @orca_build I understand that people are gonna hack ways to pty a session and sdk over it. But the prefill im not understanding advantage over just prompting directly. I have a lot of automated aliases that do this
English
@pyrons_ @JinjingLiang @orca_build I'm not getting it. -p was for background processes esp with structured output.
What is difference of --prefill and just `claude "Say hi"`
English
💡Here's a work around for the Claude Code `-p` change
This is actually a hidden option that has been in Claude Code awhile. Simply use `--prefill`
We have use it in @orca_build for months to get around the limitations
Jinjing Liang@JinjingLiang
Translation: "Congratulations, you get 25X less usage" Is my translation misleading? Yes Is their announcement misleading. YES
English
Plannotator 0.19.15 released
Plan/Annotate:
- Copyable hook path + creation guidance in Settings Hooks tab (custom plan instructions)
- (fix) Loose list continuation content now indents correctly under parent bullet
Code Review:
- jj evolution history diff mode
- Pick a specific commit as the diff base
- (fix) File comment drafts persist across close/reopen
- (fix) GitLab: concatenated JSON pagination on large MRs
- (fix) GitLab: unposted inline comments saved locally on API timeout
OpenCode:
- (fix) Commands intercepted before LLM, preventing large files from blowing up agent context
Codex:
- Hooks feature flag updated to match current CLI
- Install script now includes Codex setup guidance
Other:
- (fix) PLANNOTATOR_PORT=0 accepted without spurious warning
- (integrity) CI id-token:write scoped to AWS OIDC jobs only
English
plannotator retweetledi
planned out and ran a low stakes overnight goal to create psuedo-lsp capabilities, should I ship it?
English
x.com/plannotator/st…
^ auto create instructions for better planning using /plannotator-compound
Enable hooks in Settings > Hooks (currently only supported for planning agents)
plannotator@plannotator
1/3: compound planning skill can now introduce a hook that injects instructions based on your insights (on EnterPlanMode). This is a continuous improvement in action.
English
Plannotator 0.19.14 released today
Plan/Annotate:
- Visual Explainer skill + HTML render-annotate mode
- Code-file line range (`file.ts:10-20`) hover previews
- Hooks: let agent know about GFM syntax and allow you to customize instructions (tell your agent how to plan).
- Rendering optimization (fewer re-renders, Sonner toast migration)
Code Review:
- Configure tab size
- Configure diff line background intensity
- Ask AI now available in all-files diff view
- jj default target resolves dynamically from `trunk()`
OpenCode:
- Planning agents with display names or special characters now resolve correctly
English
@pyrons_ @ctatedev Looks cool! Will explore more & try to incorporate.
I also how some thoughts, and spec-ish, on how markdown for engineers should be rendered
x.com/plannotator/st…
plannotator@plannotator
Excited for better review surfaces. However, you can still get a lot out of markdown - whether it's from agent messages, plans, or prose-heavy docs/specs. The right rendering engine can empower markdown on top of your codebase & integrate directly with agents; without cluttering markdown or increasing the cost of tokens. In < 1min, see how plannotator.ai does it in a planning session with pi.dev
English
New proposal: mdxg.org
Markdown Experience Guidelines
A spec for how interfaces should present Markdown
Virtual pages, navigation, search, theming
Any .md file, zero changes
Reference implementation below
Would love your input
English
@0xSero Where can one find this heralded visual-explainer skill?
English
visual-explainer skill does so much heavy lifting here.
Andrej Karpathy@karpathy
This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also had some success asking the LLM to present its output as slideshows, etc. More generally, imo audio is the human-preferred input to AIs but vision (images/animations/video) is the preferred output from them. Around a ~third of our brains are a massively parallel processor dedicated to vision, it is the 10-lane superhighway of information into brain. As AI improves, I think we'll see a progression that takes advantage: 1) raw text (hard/effortful to read) 2) markdown (bold, italic, headings, tables, a bit easier on the eyes) <-- current default 3) HTML (still procedural with underlying code, but a lot more flexibility on the graphics, layout, even interactivity) <-- early but forming new good default ...4,5,6,... n) interactive neural videos/simulations Imo the extrapolation (though the technology doesn't exist just yet) ends in some kind of interactive videos generated directly by a diffusion neural net. Many open questions as to how exact/procedural "Software 1.0" artifacts (e.g. interactive simulations) may be woven together with neural artifacts (diffusion grids), but generally something in the direction of the recently viral x.com/zan2434/status… There are also improvements necessary and pending at the input. Audio nor text nor video alone are not enough, e.g. I feel a need to point/gesture to things on the screen, similar to all the things you would do with a person physically next to you and your computer screen. TLDR The input/output mind meld between humans and AIs is ongoing and there is a lot of work to do and significant progress to be made, way before jumping all the way into neuralink-esque BCIs and all that. For what's worth exploring at the current stage, hot tip try ask for HTML.
English