View raw Markdown
---
name: migrate-to-rust
metadata:
version: "1.0.0"
category: ApplicationDevelopment
description: >-
Clones, analyzes, plans, and migrates legacy codebases (C++, Fortran, Java,
Python, C, Shell) from a GitHub URL or local directory to a modern,
high-performance, deterministic, 100% Safe Rust architecture, tracking live
telemetry across every task bucket, test failure/fix cycle, and
Profile-Guided Optimization (PGO) stage, and producing a first-person AI
storytelling report and interactive HTML infographic. Use when migrating a
GitHub repository or local codebase to Rust, porting legacy C++, Fortran,
Python, or Java workloads to memory-safe Rust, designing Struct-of-Arrays
(SoA) or SIMD Rust architectures, or optimizing Rust ports with LLVM PGO.
Don't use for unrelated non-Rust refactoring or single-line syntax questions.
---
# Migrate Legacy Codebases to Safe Rust
Act as a Systems Architect, Performance Engineer, and Rust Specialist. Your goal
is to migrate legacy codebases into 100% Safe, PGO-tuned Rust while recording
measured telemetry and a first-person engineering narrative via
[`scripts/telemetry.py`](scripts/telemetry.py).
## Execution Contract & Portability
When migrating a repository or local directory (`{github_url}` or
`{codebase_path}`), follow these execution rules across agentic harnesses:
1. **Stage 0 External Code Confirmation Gate**: After cloning an external
repository and inspecting its build files, **present the resolved legacy
build command, run command, and target directory to the user and obtain
confirmation before compiling or executing cloned code**. (For local code
that the user explicitly instructed you to run, or once Stage 0 confirmation
is granted, proceed directly.)
2. **Continuous Stage 1 -> Stage 2 Progression**: Once Stage 0 baseline
capture is complete, proceed directly from writing the Architectural
Blueprint (Stage 1) into implementing and verifying the Rust workspace
(Stage 2) without pausing between milestones, where the host environment
permits autonomous tool execution.
3. **Bounded Automated Self-Healing (Max 5 Iterations)**: When tests, `clippy`
lints, or A/B parity checks fail, log the failure via
[`scripts/telemetry.py`](scripts/telemetry.py), diagnose and fix the root
cause, and re-test automatically up to a budget of **5 self-healing
iterations** (or 15 minutes per phase). If any test or seed still diverges
after the budget is exhausted, record the partial state in
`migration_telemetry.json` and document the remaining gap in the final
report rather than looping indefinitely.
4. **Verification Targets & Definition of Done**:
- `#![forbid(unsafe_code)]` Rust crate compiles with a target of `0`
`cargo clippy` warnings (any unavoidable exception must be documented).
- Unit, integration, doc, and property-based (`proptest`) tests pass with
a target of `>= 95%` branch coverage (any unavoidable exception must be
documented), with every run logged in `migration_telemetry.json`.
- A/B differential output parity is verified across `N` deterministic
seeds within a documented numerical epsilon (`--max-abs-error`), with
any floating-point/FMA/PRNG divergence root-caused and logged.
- All 4 benchmark stages (`legacy_baseline`, `safe_rust`, `soa_simd`,
`pgo_lto`) are measured and logged (or omitted cleanly if a tool such as
`llvm-profdata` is unavailable).
- The interactive HTML infographic (`migration_infographic.html`) is
rendered via `scripts/telemetry.py render-infographic`.
## End-to-End Workflow Checklist
Copy this checklist to track progress across the three stages (detailed command
recipes and architectural specifications are in
[references/pgo_and_telemetry_guide.md](references/pgo_and_telemetry_guide.md)):
```markdown
- [ ] Stage 0: Acquire repo, initialize `migration_telemetry.json` (`discovery` bucket), scan legacy codebase, confirm resolved build/run commands, and benchmark baseline
- [ ] Stage 1: Audit legacy code (`blueprint` bucket), log opportunities & latent legacy bugs, and generate 6-section Architectural Blueprint
- [ ] Stage 2.1: Scaffold `#![forbid(unsafe_code)]` Rust workspace & deterministic PRNG parity layer (`rewrite` bucket)
- [ ] Stage 2.2: Implement core domain types & SoA/SIMD kernels (<= 400 LOC/file target, full rustdoc) & benchmark `safe_rust` + `soa_simd`
- [ ] Stage 2.3: Run Verification Crucible (`test_heal` bucket, max 5 iterations) — log initial test failure % and each self-healing fix
- [ ] Stage 2.4: Run A/B differential parity harness (`parity` bucket, max 5 iterations) across N seeds within documented epsilon tolerance
- [ ] Stage 2.5: Run LLVM Profile-Guided Optimization (`pgo_bench` bucket), scan final Rust codebase, and render `migration_infographic.html`
```
--------------------------------------------------------------------------------
## Stage 0: Repository Acquisition, Telemetry Init & Legacy Baseline
Resolve context parameters from the repository or directory:
Parameter | Placeholder | Resolution Strategy
:-------------------------------- | :----------------------------------------- | :------------------
**Source Input** | `{codebase_path}` or `{github_url}` | Accept a GitHub URL (`https://github.com/...`) or a local filesystem path
**Original Language / Framework** | `{legacy_language}` | Auto-detect from file extensions and build files (`CMakeLists.txt`, `Makefile`, `meson.build`, `pyproject.toml`, `pom.xml`, `*.f90`, `*.cpp`)
**Baseline Run Command** | `{baseline_run_command}` | Identify from `README.md`, CI workflows (`.github/workflows/*.yml`), `Makefile`, or test scripts
**Telemetry Ledger** | `{workspace_dir}/migration_telemetry.json` | Initialized via `scripts/telemetry.py init` at the start of Stage 0
### Stage 0 Execution Steps
1. **Clone Repository (if GitHub URL)**: Choose a clean local directory
(default to `~/rust_migrations/{repo_name}`, or inside the current workspace
if requested) and run `git clone {github_url}
~/rust_migrations/{repo_name}`.
2. **Initialize Telemetry & Scan Legacy Codebase**: Run `scripts/telemetry.py
init --telemetry "{workspace_dir}/migration_telemetry.json" --project-name
"{project_name}" --legacy-lang "{legacy_language}"`. Start the `discovery`
bucket and run `scripts/telemetry.py scan-codebase --phase before` on the
legacy source directory to record baseline file count, code LOC, comment
density %, max file LOC, nesting depth, average function length, and raw
pointer/memory hazards.
3. **Confirm & Benchmark Legacy Baseline**: Before executing build or run
commands from a cloned external repository, surface the resolved build
command, `{baseline_run_command}`, and working directory for user
confirmation. Once confirmed, compile the legacy project in release mode
(`-O3` / `Release`), capture golden reference outputs across deterministic
seeds, measure baseline wall-clock runtime (ms) and peak RSS (MB), and log
Stage 1 via `scripts/telemetry.py log-benchmark --stage legacy_baseline`.
--------------------------------------------------------------------------------
## Stage 1: Deep Audit, Opportunity Discovery & Architectural Blueprint
Start the `blueprint` timer bucket (`scripts/telemetry.py start-bucket --bucket
blueprint`), inspect the repository, log every discovered opportunity and
**latent legacy bug** (`scripts/telemetry.py log-opportunity`), and generate the
6-section Markdown blueprint defined in Section 2 of
[references/pgo_and_telemetry_guide.md](references/pgo_and_telemetry_guide.md).
**Stage 1 -> Stage 2 Transition**: Save the blueprint and proceed directly to
Stage 2 (unless the user explicitly requested a blueprint-only plan).
--------------------------------------------------------------------------------
## Stage 2: Phased Safe Rust Implementation, Self-Healing, PGO & Infographic
Follow the command recipes and core engineering standards in
[references/pgo_and_telemetry_guide.md](references/pgo_and_telemetry_guide.md)
for each step:
### Step 1: Scaffold Safe Rust Workspace & Port Modules (`rewrite` Bucket)
1. Start the `rewrite` timer bucket (`scripts/telemetry.py start-bucket
--bucket rewrite`).
2. Initialize the Rust workspace with `#![forbid(unsafe_code)]` in `src/lib.rs`
and `src/main.rs`.
3. Implement the deterministic PRNG wrapper, domain types, and modules in
dependency order (`<= 400` LOC per file target, complete `//!` and `///`
rustdoc with runnable doc-tests).
4. Benchmark and log Stage 2 (`safe_rust`) and Stage 3 (`soa_simd`) via
`scripts/telemetry.py log-benchmark`.
### Step 2: The Verification Crucible — Test, Fail, Diagnose & Fix (`test_heal` Bucket)
1. Start the `test_heal` timer bucket (`scripts/telemetry.py start-bucket
--bucket test_heal`).
2. Write unit tests, runnable doc-tests, and `proptest` boundary tests.
3. Run `cargo test --all-targets`, `cargo test --doc`, and `cargo clippy
--all-targets`.
4. **Log Every Iteration (Up to 5 Iterations)**: After Run 1 and each
subsequent run, invoke `scripts/telemetry.py log-test-run` with total tests,
passed, failed, coverage %, clippy warnings, diagnostic summary, and fix
applied. Iterate until `0` failures and `0` warnings (or until the
5-iteration budget is reached, documenting any unavoidable exception).
### Step 3: A/B Differential Output Parity (`parity` Bucket)
1. Start the `parity` timer bucket (`scripts/telemetry.py start-bucket --bucket
parity`).
2. Run the automated A/B differential harness across `N` deterministic seeds
comparing `{baseline_run_command}` and the Rust binary within the documented
epsilon tolerance.
3. Log each parity iteration via `scripts/telemetry.py log-parity-run`. If any
seed diverges beyond epsilon, bisect the first diverging step, repair the
discrepancy, and re-run (up to 5 iterations).
### Step 4: Profile-Guided Optimization (PGO) & Final Benchmarks (`pgo_bench` Bucket)
1. Start the `pgo_bench` timer bucket (`scripts/telemetry.py start-bucket
--bucket pgo_bench`).
2. Execute the LLVM PGO workflow in
[references/pgo_and_telemetry_guide.md](references/pgo_and_telemetry_guide.md):
build with `RUSTFLAGS="-Cprofile-generate=./pgo-data"`, run representative
workloads, merge profiles via `llvm-profdata`, and rebuild with
`RUSTFLAGS="-Cprofile-use=./pgo-data/merged.profdata"` and `lto = "fat"`
(or use the documented `-Ctarget-cpu=native` fallback if `llvm-profdata` is
unavailable).
3. Verify that tests and A/B parity still pass on the PGO binary, then measure
wall-clock runtime and peak RSS and log Stage 4 (`pgo_lto`) via
`scripts/telemetry.py log-benchmark --stage pgo_lto`.
### Step 5: Final Census, First-Person AI Story & Visual Infographic
1. Run `scripts/telemetry.py scan-codebase --phase after` on `src/` to capture
final LOC, `<= 400` LOC compliance, comment/rustdoc density, runnable
doc-tests, nesting depth, and `unsafe` block count.
2. Log a **Before vs. After Code Spotlight** (`scripts/telemetry.py
log-spotlight`) and populate the **First-Person AI Story Chapters**
(`scripts/telemetry.py set-story`).
3. End the `pgo_bench` timer bucket and render the interactive HTML infographic
(where `{skill_dir}` is the directory containing this `SKILL.md`):
```bash
python3 "{skill_dir}/scripts/telemetry.py" render-infographic \
--telemetry "{workspace_dir}/migration_telemetry.json" \
--out "{output_dir}/migration_infographic.html" \
--scope-summary "{short_summary_of_ported_modules}"
```
4. Output a concise executive summary referencing the generated
`migration_infographic.html` report and key measured before/after metrics.
--------------------------------------------------------------------------------
## Bundled Resources
- **Telemetry & Infographic CLI**:
[`scripts/telemetry.py`](scripts/telemetry.py)
- **Engineering Standards, Blueprint Spec, PGO & Telemetry Recipes**:
[references/pgo_and_telemetry_guide.md](references/pgo_and_telemetry_guide.md)
- **Interactive HTML Infographic Template**:
[references/infographic_template.html](references/infographic_template.html)
