panto-subagents.git/subagents, branch main

Show full subagents.lua source when expanded; guide toward templated prompts

2026-09-03T13:57:03Z

The expanded progress board capped the executed Lua at 128 lines (and 32KB), so long workflow sources ended in an ellipsis even after Ctrl-O. Expanded now renders the whole source; collapsed still shows the first four lines. Add a "Program, don't transcribe" section to the system guidance so the primary writes shorter sources: one [[template]] with {name} placeholders filled by gsub with a table, loops over data instead of per-child copies, shared system prompts in `agents`, and file references instead of quoted contents. A single templating mechanism is named on purpose to remove choice. Co-Authored-By: Claude Fable 5.1 Claude-Session: https://claude.ai/code/session_01ETqhv5FVD8Z14BEBsXeHHC

Defer workflow completion wake until callback unwinds

2026-09-02T16:47:37Z

Sort completed subagents above active ones

2026-09-02T15:30:55Z

Expose background workflow activity events

2026-09-02T14:26:58Z

Bracket each background workflow with lifecycle events and emit its primary wake before completion so integrations do not report a premature idle state.

Show executed Lua in subagent progress boards

2026-09-02T14:26:44Z

Render subagents.lua source in both collapsed and expanded progress views so inspection-only calls remain understandable and survive session replay.

Use Panto terminal width helpers for progress rendering

2026-08-26T15:17:36Z

Keep subagent boards pinned while background workflows run

2026-08-21T20:12:35Z

A background subagents.lua workflow outlives the tool call that started it, so settling the outer result unpinned a board whose agents were still working. Track a per-tool-call workflow count and derive the pin from both the tool being active and that count, so the board only returns to transcript order once its last workflow finishes. Also show the reasoning level alongside the model on agent cards.

Add workflow cancellation controls

2026-08-21T17:26:48Z

Expose a json codec to the subagents.lua sandbox

2026-08-20T21:32:46Z

Model-authored workflow Lua could reach string, table, math, and utf8 but had no way to parse an agent's report, so any fan-out over discovered work had to invent an ad-hoc line format and hand-roll string.gmatch. Hand the guest panto.ext.json instead. decode returns nil plus a message rather than raising: the sandbox deliberately has no pcall, so a raising decoder would let one malformed agent line kill an entire workflow. Teach the pattern in the injected guidance too: deterministic Lua owns the loops, counts, ordering, and gates, agents supply discovery and judgment, and a worker that feeds Lua ends its report with one parseable line. The existing static-graph example stays; a discover-then-fan-out example joins it.

Replace lyaml with api7-lua-tinyyaml and prepare the 0.1.0 release

2026-08-20T19:39:46Z

lyaml binds the system libyaml, which luarocks does not vendor, so installing the rock failed wherever that library was absent, taking the whole extension down on a stock macOS machine. api7-lua-tinyyaml is pure Lua and matches lyaml on every quoting, escaping and block-scalar form profiles use. What the subset costs: anchors and aliases are not resolved, and plain multi-line scalars raise. The raise surfaces through the existing parse warning, but an unresolved alias came back as the literal "*alias" string and would have become a profile name, so frontmatter.parse now detects those keys, drops them, and warns. Point source.url at code.tjp.lol with a v0.1.0 tag, replacing a GitHub URL that never existed. Add a LICENSE file and mise release tasks, and fix the README dependency-install command, which called a luarocks.cmd function that does not exist.