tkersey/dotfiles

Run a v2-only Node JSONL proxy that spawns `codex app-server` and exposes an automation-friendly stream API. Use when you need to drive the app-server programmatically (automation/orchestration/session mining), have subagents send updates or code patches into live sessions, auto-handle approvals, forward server requests with deterministic timeouts, mine sessions via `thread/*`, steer active turns (`turn/steer`), or run N parallel instances (each instance is one proxy + one app-server child).

Run a v2-only Node JSONL proxy that spawns `codex app-server` and exposes an automation-friendly stream API. Use when you need to drive the app-server programmatically (automation/orchestration/session mining), have subagents send updates or code patches into live sessions, auto-handle approvals, forward server requests with deterministic timeouts, mine sessions via `thread/*`, steer active turns (`turn/steer`), or run N parallel instances (each instance is one proxy + one app-server child).

Review+fix protocol with safety guardrails (unsoundness, invariants, footguns, incidental complexity). Use when prompts say "$fix this PR", "fix current branch", "fix this diff", "repair CI red", or "apply a minimal patch", and when crash/corruption/invariant-break issues need correction with a validation signal.

Create a language-agnostic ghost-library package from an existing library repo by extracting SPEC.md, exhaustive tests.yaml, INSTALL.md, README.md, VERIFY.md, and upstream LICENSE files with provenance and regeneration instructions. Use when prompts say "$ghost", "ghostify this repo", "spec-ify/spec-package this library", "ghost library", or ask to extract portable spec/tests from source tests, including stateful/protocol workflows such as multi-step agent loops; do not use for implementation work or editing skills.

Clarify ambiguous or conflicting requests by researching first, then asking only judgment calls. Use when prompts say "$grill-me"/"grill me", "ask hard questions", "pressure-test assumptions", "clarify scope/requirements", "define success criteria", or request system-design/optimization decisions before implementation; stop before implementation.

Capture and persist evidence-backed execution learnings into `.learnings.jsonl`. Use when a coding slice completes or pauses, after failing-to-passing validation transitions, repeated failure loops, pivots/retries/reversals, footgun discovery, or before commit/PR handoff. Distill actionable do-more/do-less rules with concrete evidence and application context so future turns reuse what works and avoid known traps.

Swarm coordinator activated only by explicit `$mesh` invocation. Trigger cues/keywords: orchestrate subagents/workers, execute ready `$st` slices/tasks, run propose/critique/synthesize/vote cycles, coordinate concurrent workers, require consensus + validation before completion, and persist learnings via `$learnings`.

Create micro-patches from staged git changes (minimal incision) with at least one validation signal per patch. Use when asked to split work into small .patch files, export/share diffs, or produce patches instead of commits.

Launch and manage Codex Cloud tasks from the CLI, including detached background watchers that track completion. Use when users ask to run coding work in cloud/background agents, queue multiple cloud tasks, poll task status, fetch cloud diffs, apply cloud outputs locally, or pair cloud kickoff with `$cas` orchestration.

Deconstruct high-cost abstractions and recommend lower-level primitives that reduce tooling, indirection, and hidden steps. Use when requests ask for fewer layers (for example "too many layers", "remove this framework/plugin/DI", "ditch codegen/task runners", "replace with plain scripts/config/SQL"), or ask for an abstraction audit with a prioritized cut list and phased migration with rollback (analysis-only unless implementation is explicitly requested).

Manage persistent task plans in repo-committed JSONL (`.step/st-plan.jsonl`) so state survives turns/sessions and stays reviewable in git. Use when users ask to "use $st", "resume the plan", "export/import plan state", "checkpoint milestones", "track dependencies/blocked work", "show ready next tasks", "keep shared TODO status on disk", "map a `$select` plan into durable execution state", "prove `$st` works for implementation tracking", or diagnose/repair `st-plan.jsonl` concerns (for example append-only vs mutable semantics, lock-file gitignore policy, or seq/checkpoint integrity).

Software surgery protocol for proof-backed, minimal-diff implementation: Contract to Invariants to Creative Frame to inevitable incision to Proof. Use when users say "run $tk", "patch-first", "keep the diff small", "state contract + invariants", "one clear incision", "fix it with a validation signal", or orchestration cues like "workers use $tk" / "internally use $tk".

Use when requests involve Zig code or toolchain work: editing `.zig`, changing `build.zig`/`build.zig.zon`, running or fixing `zig build|test|run|fmt|fetch`, debugging compile/runtime/test failures, comptime/reflection/codegen, allocator ownership, SIMD (`std.simd`/`@Vector`), threads (`std.Thread`/`Thread.Pool`), cross-compilation, zero-copy parsing, or C interop (`@cImport`). Enforce correctness-first validation with tests, `std.testing.fuzz`, and allocation-failure checks.