Skip to content

donaldfilimon/abi / .agents/skills/run-abi

run-abi

Verified

Build, launch, and drive the abi stable-Rust project — the `abi` CLI and the `abi-mcp` JSON-RPC server. Use when asked to run, start, build, smoke-test, or screenshot abi, drive its CLI subcommands, exercise the MCP server over stdio, or confirm a change works in the real binaries (not just the test suite).

Run abi

abi is a stable Rust 1.99.1 framework that builds two binaries: a CLI (target/debug/abi) and an MCP server (target/release/abi-mcp in this smoke, JSON-RPC 2.0 over stdio plus an automatically attempted custom loopback HTTP listener). The MCP listener serves persistent MCP 2024-11-05 HTTP+SSE sessions plus a one-shot direct-reply POST /message mode; it is not Streamable HTTP. The driver builds and exercises the real binaries end-to-end.

The driver overrides inherited store settings with ABI_WDBX_PATH=:memory: and ABI_WDBX_PERSIST=0. Its explicit-path WDBX round trip uses a scratch store under target/debug/. A failed build stops before binary execution. MCP runs through mcp/launcher.sh stdio; the driver rebuilds release MCP because the launcher prefers that profile. Builds explicitly select this checkout's target/ directory, overriding an inherited CARGO_TARGET_DIR so they refresh the binaries the driver executes.

Paths below are relative to the repo root. The driver lives at .agents/skills/run-abi/smoke.sh and resolves the repo root from its own location.

Run (agent path) — the driver

./.agents/skills/run-abi/smoke.sh

Expected tail on success (exit 0):

=== summary: pass=N fail=0 ===
transcript: <repo>/target/debug/run-abi-smoke.txt
SMOKE OK

Prerequisites

  • Stable Rust 1.99.1 via rust-toolchain.toml. Always use ./tools/cargo.sh (never bare Homebrew cargo).
  • Primary gate: ./tools/check.sh.

Build

./tools/cargo.sh build --locked -p abi-cli   # -> target/debug/abi
./tools/cargo.sh build --locked --release -p abi-mcp   # launcher-preferred profile

On arm64 macOS, building with the default foundationmodels feature also produces target/debug/libabi_fm_shim.dylib next to the binaries.

Drive the CLI directly

./target/debug/abi help
./target/debug/abi backends
./target/debug/abi scheduler status
ABI_WDBX_PATH=:memory: ./target/debug/abi complete "summarize scheduler status"
./target/debug/abi plugin list

Drive the MCP server directly

printf '%s\n' \
  '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"probe","version":"0"}}}' \
  '{"jsonrpc":"2.0","id":2,"method":"tools/list"}' \
  | ABI_WDBX_PATH=:memory: ABI_WDBX_PERSIST=0 ./mcp/launcher.sh stdio

Exactly 12 frozen tools. Always use mcp/launcher.sh so binary selection and the FM dylib resolve together.

Gotchas

  • GPU honesty: accelerated=false means native kernels are not linked; CPU SIMD is the real path. Do not treat Metal-as-preferred as native kernels.
  • complete (no --live) is fully local. --live is Anthropic-only for HTTP; apple-fm --confirm uses the FoundationModels shim when ready.
  • Store safety: do not point smokes at the user's real ~/.abi/ path.

Troubleshooting

SymptomFix
build FAIL./tools/check.sh; ensure the repository toolchain via ./tools/cargo.sh
dyld FM shim missingbuild from repo so libabi_fm_shim.dylib sits next to abi
MCP emptyJSON on stdout; logs on stderr — drop stderr when grepping tools
finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the workStars: 322kUpdated Claude CodeCodexVerifiedInstall
systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixesStars: 322kUpdated Claude CodeCodexVerifiedInstall
test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation codeStars: 322kUpdated Claude CodeCodexVerifiedInstall
skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.Stars: 198kUpdated Claude CodeCodexVerifiedInstall