Place the mobile agent GUI on the agent's own config channels - #275
Draft
jiweiyuan wants to merge 1 commit into
Draft
Place the mobile agent GUI on the agent's own config channels#275jiweiyuan wants to merge 1 commit into
jiweiyuan wants to merge 1 commit into
Conversation
Records the three structured channels every coding agent exposes — hooks, transcript, and a sidecar session protocol — and why Termio can only use the first two: the sidecar replaces the runtime, and the TUI is the product. Draws the data/code line for per-agent support. The ten shipped manifests already carry zero grid regexes, so the content and interaction planes add only locations and key names (`transcript`, `questions`) selected from closed sets. Parsing a transcript shape stays Swift: waku spends 300-530 lines of Rust per provider and linkcode 2155 on Claude Code alone, so no shipping implementation configures an agent as pure data. Adds linkcode and waku to the prior art, and narrows the open Codex question to whether hooks.json can write back a decision — its app-server approvals are unreachable while the TUI runs.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
jiweiyuan
marked this pull request as draft
August 13, 2026 17:32
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Revises
design/mobile-agent-ui-protocol.mdto v3, answering two questions: how the iOS GUI gets its content for Claude Code and Codex, and how much of per-agent support can live in a manifest instead of Swift.Three channels, two usable
Every coding agent exposes structure three ways — lifecycle hooks, an on-disk transcript, and a sidecar session protocol. The first two are zero-intrusion; the third (Codex app-server, Claude
stream-json, ACP) replaces the runtime, and the TUI is the product. So the content plane comes from the transcript, the interaction plane from hooks, and grid-reading survives only as the observed-level fallback already described in §6.4.Where the config line falls
The ten shipped manifests already carry no grid regexes —
titleStatusreads the OSC title, and the old screen-scrapestatusfield is unused. Keeping that property costs two new manifest blocks,transcriptandquestions, both carrying locations and key names picked from closed sets, matching howresumeandhooks.dialectalready work.Parsing a transcript shape stays Swift, and the doc says why rather than leaving it to taste: waku spends 300–530 lines of Rust per provider, linkcode 2155 lines on Claude Code alone. Neither configures an agent as pure data, so a JSONPath mapping DSL would buy a mini-interpreter that is hard to write and harder to test.
Also
hooks.jsonhas the same write-back.Docs only — no code paths touched. The
docs/README.mdindex was regenerated from front matter.Release Notes: None