Skip to content

stevesolun/ctx / docs

skill-router

Flagged

Repo-aware recommendation manager for ctx. Scans the active repository, identifies stack and workflow signals, recommends a capped set of skills, agents, and MCP servers, and unloads helpers that no longer match the current work after user confirmation. Harnesses are recommended by the custom-model onboarding flow or loop adapters and then attach to the same recommendation layer.

FileKindSize
SKILL.mdskill md9.3 KB
assets/javascripts/catalog.jsscript2.5 KB
assets/javascripts/repo-stats-refresh.jsscript3.2 KB
backup-hook-install.mdmarkdown9.2 KB
catalog.mdmarkdown10.3 KB
ctx-fit/ARCHITECTURE_CANDIDATES.mdmarkdown12.9 KB
ctx-fit/ARCHITECTURE_GAP_ANALYSIS.mdmarkdown10.7 KB
ctx-fit/CTX_AUDIT.mdmarkdown11.6 KB
ctx-fit/CURRENT_STATE.mdmarkdown16.3 KB
ctx-fit/DECISIONS.mdmarkdown15.7 KB
ctx-fit/GOOD_FIRST_ISSUES.mdmarkdown20.6 KB
ctx-fit/IMPLEMENTATION_TRACKER.mdmarkdown697 B
ctx-fit/MAP.mdmarkdown7.8 KB
ctx-fit/PRD.mdmarkdown9.7 KB
ctx-fit/STATE.mdmarkdown60.9 KB
ctx-fit/milestones/M1_SPIKE_FRICTION_LOG.mdmarkdown5.5 KB
ctx-fit/milestones/MILESTONE_00_AUDIT.mdmarkdown5.1 KB
ctx-fit/planned_tasks.mdmarkdown17.7 KB
dashboard.mdmarkdown24.7 KB
enterprise-readiness-review.mdmarkdown8.7 KB
entity-onboarding.mdmarkdown18.0 KB
harness/attaching-to-hosts.mdmarkdown16.2 KB
harness/clean-host-contract.mdmarkdown4.5 KB
harness/loopflow-adapter-demo.mdmarkdown11.1 KB
huggingface-publish.mdmarkdown3.3 KB
index.mdmarkdown22.4 KB
knowledge-graph.mdmarkdown22.7 KB
marketplace-registry.mdmarkdown4.0 KB
memory-anchor.mdmarkdown2.7 KB
plans/ctx-agent-reviewer-workbench.mdmarkdown12.4 KB
plans/unified-capability-engine.mdmarkdown59.0 KB
qa/dashboard-user-story-status.csvother26.0 KB
qa/feature-user-story-status.csvother92.5 KB
research/context-engineering-guidance-audit.mdmarkdown10.4 KB
services/macos/com.claude.backup.watchdog.plistother1.7 KB
services/systemd/claude-backup-watchdog.serviceother1.3 KB
skill-lifecycle-and-dashboard.mdmarkdown6.2 KB
skill-quality-install.mdmarkdown6.2 KB
skill-router/index.mdmarkdown2.5 KB
skill-stack-matrix.mdmarkdown5.5 KB
skills-health.mdmarkdown3.5 KB
stack-signatures.mdmarkdown6.2 KB
telemetry.mdmarkdown24.7 KB
threat-model.mdmarkdown15.0 KB
toolbox/behavior-miner.mdmarkdown2.6 KB
toolbox/configuration.mdmarkdown4.0 KB
toolbox/council-runner.mdmarkdown2.4 KB
toolbox/hooks.mdmarkdown2.6 KB
toolbox/index.mdmarkdown3.0 KB
toolbox/intent-interview.mdmarkdown3.0 KB
toolbox/starters.mdmarkdown3.0 KB
toolbox/templates/docs-review.jsonconfig599 B
toolbox/templates/fresh-repo-init.jsonconfig533 B
toolbox/templates/refactor-safety.jsonconfig634 B
toolbox/templates/security-sweep.jsonconfig636 B
toolbox/templates/ship-it.jsonconfig676 B
toolbox/verdicts.mdmarkdown4.0 KB

ctx-fit/milestones/MILESTONE_00_AUDIT.md

View on GitHub
# Milestone 0 — Audit · Milestone Report

Date: 2026-08-09 · Branch `ctx-fit/m0-baseline` · Baseline commit `e8af9554`

**Gate outcome: ACCEPTED WITH CONDITIONS — all conditions met.**

Both reviewers returned `accept-with-conditions`. Every blocking objection was
fixed rather than argued. This report is for a reader who does not know CTX
internals.

## What a user can do now that they could not before

Point one command at any repository and get, free, read-only, in under a
second:

- the repository's own test, lint, type-check and build commands, each with the
  exact manifest stanza that justified it and a coarse confidence;
- an inventory of the AI coding setup already present (`AGENTS.md`,
  `CLAUDE.md`, tool config, installed skills);
- a plain verdict on whether the repository can be evaluated at all, and if
  not, what to fix first.

```bash
pip install claude-ctx && cd my-project && ctx fit
```

## Why that is valuable

It replaces a manual pass over `pyproject.toml`, `package.json`, `go.mod` and
`Makefile` — and, more importantly, it is honest. The previous surface
(`ctx-scan-repo --recommend`) inferred relevance from keyword overlap and wrote
into the user's home directory. `ctx fit` writes nothing, guesses nothing, and
cites the file and stanza behind every claim.

## What the gate caught, and what changed

The reviews found real defects. They are listed because the process working is
the point of this milestone.

| Finding | Severity | Resolution |
| --- | --- | --- |
| **Evaluability overclaim.** A repository with a pytest stanza and zero test files was reported as "can be evaluated: it has deterministic tests". A declared runner is intent, not evidence. | **Blocking** — this was the one sentence the product's credibility rests on | `has_deterministic_verification` now requires a declared command **and** observed test material. The repository is told exactly why it is not evaluable and what to do. Two regression tests added. |
| **Non-deterministic output.** `--json` embedded a wall-clock timestamp and an absolute path from the legacy scanner, so two consecutive runs of a "versioned, reproducible" schema differed. | **Blocking** | Volatile fields stripped; a test now asserts byte-identical serialization across runs. |
| **Audit headline overstated.** "93% of the engine lane is removable" conflated *not reachable* with *not needed*. Only 10,741 LOC is proven dead; 35,993 LOC is reachable today. | **Blocking** | `CTX_AUDIT.md` §3 rewritten to separate the two claims. The deletion program starts with the proven-dead cluster only. |
| **Three contradictory execution decisions live at once** across `MAP.md`, `ARCHITECTURE_GAP_ANALYSIS.md` and `DECISIONS.md`. | **Blocking** | `DECISIONS.md` declared authoritative; superseded notices added to `MAP.md`. |
| **Two contradictory task ledgers.** | **Blocking** | `IMPLEMENTATION_TRACKER.md` retired; `planned_tasks.md` is the single board. |
| **Undiscoverable at the front door.** `ctx fit` appeared nowhere in the README; `ctx --help` still described a "model-agnostic harness". | **Blocking** | README now leads with the product and real output; parser description rewritten. |
| **Extraction estimate ~2x optimistic; delete list swallowed the contamination controls the PRD depends on.** | Correction | Estimate revised; ~6,012 LOC of holdout/exposure-ledger code reclassified ADAPT. |
| CTX-internal jargon in user output; a permanent `[no] coding-harness` row. | Simplification | Collapsed to one plain sentence; full detail kept in `--json`. |

## The most consequential finding

An AST import graph from every real entry point proved that **10,741 LOC of the
event-sourced capability lifecycle has no production consumer** — it is
reachable only from its own tests. An independent second review reproduced this
from a broader seed set and concluded the original audit *understated* it.

That cluster includes work completed earlier in the same session. It was
correct engineering against the previous goal; the goal changed. Recorded as
ADR-010 rather than quietly protected.

## What evidence proves this works

- 14 tests in `src/tests/fit/`, including adversarial cases: no tests, declared
  runner with no tests, malformed `pyproject.toml` and `package.json`, empty
  repository, missing path, Node lockfile runner selection.
- `mypy` clean across 554 source files; Ruff and formatting clean.
- The full 19-lane PR preflight was green at the baseline commit: 8,233 tests,
  91.9 percent coverage.
- The reviewer's exact repro of the overclaim now produces the correct refusal.

## What still prevents charging money

The product currently answers *"can this repository be evaluated, and what
should I fix?"* It does not yet answer *"which setup should I use?"* — there is
no readiness score, no candidate set, no experiment, and no recommendation.
Roughly the first third of the first clause of the promise is implemented.

## Next smallest step that increases usefulness

**M3 — AI Agent Readiness.** It converts the profile from information into a
prioritized decision, needs no provider spend, and is the last milestone before
candidate generation. The design is ready.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall