shiro-neko/TODO.md

TODO

Next up. One item, one outcome, verifiable when done.

Longer-term direction lives in ROADMAP.md.

---

Now

Summarize the pruned span

Compaction now keeps the model's memory of a turn, but it still tells the model nothing about the messages it dropped, so a decision from forty messages ago can be contradicted with confidence.

• Summarize the discarded messages before dropping them (prune.droppedSpan + session.summarizeDiscarded)
• Inject the summary in place of the count (Note (retained from compacted history) appended to history)
• Budget it: a summary that grows with the session defeats the point (6k excerpt + 3-6 lines, one call per compaction)
• Test: a pruned decision is still recoverable from the summary (test/compact.test.ts lossless suite)

Hot-reload an installed entry

/registry add writes the file and says to restart. The skill catalogue and the guard chain are both assembled at boot, so a mid-session install does nothing until then.

• Rebuild the skill list and plugin host after an install or removal (Session.updateSkills/updatePlugins + cli.tsx hot-reload)
• Leave a turn in flight alone: its rules must not change underneath it (pendingSkills/pendingHost + drainPendingHotReload at turn boundary)
• Test: a skill installed mid-session is callable in the next turn without a restart (test/hot-reload.test.ts)

---

Next

MCP without the schema tax

Every MCP tool's schema goes into the prompt today, so twenty tools from one server cost roughly 2,750 tokens per request whether the model uses them or not. toolSets does not gate them.

phi solves this with three meta-tools — mcp_list, mcp_inspect, mcp_call — and a prompt that names only the servers. A hundred servers then cost almost nothing until one is called.

• mcp_list / mcp_inspect / mcp_call replacing per-tool registration
• The prompt lists server names, not schemas
• Calls go through the same permission rules and guard as a built-in
• Keep per-tool registration as an option: a two-tool server is cheaper registered directly
• Test: a configured server contributes no schema to the request until mcp_call

Derive the tool-name lists

TOOL_SETS and MUTATING_TOOLS both list names by hand. A tool added to one and forgotten in the other is a silently ungated write, which is the worst kind of bug this codebase can have.

• Mark each tool as mutating where it is defined, not in a list beside it
• TOOL_SETS covers every registered tool, checked rather than assumed
• Test: a tool in no set, or a mutating tool outside MUTATING_TOOLS, fails the suite

Subagent parallelism

Two independent searches run sequentially. The panel already renders several agents; the loop does not fan out.

• task accepts several investigations and runs them together
• Test: two delegated searches overlap in time rather than queueing

Undo a turn

Every comparable CLI has this: opencode /undo and /redo, Claude Code /rewind with checkpoints. There is /resume here, which restores a session, and nothing that walks one back.

• Snapshot files before each prompt, capped at the 100 most recent
• /undo restores files, conversation, or both; /redo reverses it
• Say plainly what is not covered: a bash command's effects cannot be snapshotted
• Test: an edit is reverted, and the model's own record of it goes with it

---

Maintenance

• Pricing table needs a source note and a date; rates drift and ours are hand-entered
• estimateTokens divides JSON length by four. Good enough for a compaction threshold, wrong enough to mislead in /cost. Either label it an estimate everywhere or use a real tokenizer
• listPaths walks up to 5000 files once per session. Fine for a repo, wasteful in a monorepo, and it never notices a file created after the first @
• MUTATING_TOOLS is now only used by tests and docs; the permission defaults are what actually gate a write. Either delete it or make the defaults derive from it

---

Known rough edges

Not bugs exactly, but things that will bite someone.

• /clear wipes the terminal scrollback. <Static> output is already committed, so clearing React state alone leaves it on screen. The escape sequence works but takes the user's earlier terminal history with it.
• Memory has no conflict resolution. Two contradictory notes both persist and both get injected. /memory may merge them, or may keep both.
• Windows cmd /c differs from bash -lc. A command the model writes for one shell may fail on the other. The prompt states the platform; it does not translate.
• An unknown name in toolSets is dropped silently. The header line shows which sets actually loaded, but a typo reads as "that set is off" rather than as a mistake.
• Permission rules gate the call, not what it does. bash with git * allowed will run a git alias that shells out to anything, and there is no sandbox around the shell. Codex solves this with OS-level isolation — Seatbelt, Landlock, a Windows equivalent — which is three platform-specific implementations and not something to half-ship.
• The reasoning panel is per-turn, not per-step. Reasoning from an early step stays on screen through later ones until the turn ends.
• An interrupted command's effects are unknown, and the model is told so. Nothing can know how far a half-run migration got.
• @ completion lists files, not directories. @src/ narrows correctly, but you cannot complete to src/ itself, because the walker only yields files.
• An installed skill is a stranger's words in your system prompt. The install shows the body first and /skills records the origin, but nothing re-checks it later: a registry that changes a URL's contents affects the next install, not one already on disk.
• A registry index is trusted for its contents, not its authorship. There are no signatures. registryUrl is the whole trust decision.

---

Done

Kept for one release, then deleted. The 1.0.0 release batch:

• A spend ceiling (maxSpendUsd): checked before each turn, refused at 100% naming the ceiling, warns once at 80%, headless exits non-zero. Unpriced models are not enforced
• A cheaper subagent model (subagentModel): explore resolves against it, review and worker keep the parent's, /cost splits subagent spend by model id
• Twenty new built-in tools (41 total) in a new extra set: line edits, filesystem navigation, read-only git extensions, and code/environment reads
• Twenty new bundled skills (29 total) plus the eleven originals deepened; all moved to src/skills-md/*.md as the Markdown source of truth, embedded at build
• Ten new data-only plugins: safety refusals on by default (force push, pipe-to-shell, root, env credential writes) and opt-in workflow plugins (conventional commit, tests-first, small diffs, main-branch commits, git config, confirm-delete)
• Custom slash commands from Markdown files, with $ARGUMENTS/$1 and guarded shell substitution; a custom command never shadows a built-in
• Auto-loaded external skills, tools, and plugins from ~/.shiro-neko/<kind> and .shiro/<kind>, all data, never code; a bad file is reported and skipped
• The welcome interface redesigned into a structured dashboard with a session banner, a grouped environment panel, and a meta bar; the input in a two-tone box with a split footer
• The system prompt advanced: a failure-recovery loop, a delegation policy, compaction awareness
• The release workflow's dead dry_run input wired: manual dispatch publishes only when unchecked, tag pushes always publish