stevesolun/ctx / docs
skill-router
FlaggedRepo-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.
enterprise-readiness-review.md
View on GitHub# Enterprise Readiness Review
Status date: 2026-07-26
Repository evidence was reviewed through commit
`8d2a91ef65afb8bec0c8da2169fd174b444da22d`; this report is introduced by the
subsequent documentation commit. GitHub-hosted controls are identified
separately because they are not established by repository files.
This review checks the user-provided enterprise-readiness task packet against
the current repository. It is an evidence record, not a promise that an active
patch or external control has shipped.
## Product Decision
ctx currently supports a local, single-user CLI and monitor. The shared-service
model described in the packet is not the current deployment contract.
Container deployment, OIDC/SSO, multi-user RBAC, and service isolation remain
conditional work until that product model is explicitly adopted.
Enterprise work for the current model prioritizes:
1. private security reporting and repository governance;
2. dependency, static-analysis, and release-supply-chain controls;
3. source and license provenance;
4. local monitor trust boundaries;
5. truthful telemetry and benchmark evidence;
6. reproducible packaging and maintainability.
## Checked Status
| Packet item | Status | Repository evidence | Remaining obligation |
|---|---|---|---|
| P0-1 Security disclosure | Partial | `SECURITY.md`; GitHub private vulnerability reporting enabled | Decide and document response targets, or formally amend that packet criterion; revalidate the repository setting during the release audit |
| P0-2 Independent ownership | External/human blocker | `.github/CODEOWNERS` exists but names only one human owner | Add a second human owner |
| P0-2 Required review controls | Partial; external/human blocker | Active GitHub [ruleset 15907020](https://github.com/stevesolun/ctx/rules/15907020) targets `main`, strictly requires the `CI required` status, and has no bypass actors | Add a second human CODEOWNER, then require pull-request review and CODEOWNER approval and validate those controls with a test PR |
| P0-3 Dependabot, CodeQL, dependency audit | Implemented locally | `.github/dependabot.yml`, CodeQL and dependency-audit workflows, hash-locked policy dependencies, fail-closed Dependabot parser, and security-config classifiers | Pass the final remote workflow matrix and observe the first scheduled Dependabot/audit runs |
| P0-4 License provenance | Critical partial | `src/ctx/core/source_registry.py`, `src/import_designdotmd_skills.py`, generated `NOTICE`, and paired adversarial tests | Extend the enforced registry boundary to every ingestion/recommend/install/publication path; keep unresolved full bodies blocked |
| P1-1 Local monitor trust model | Implemented with accepted substitution | Exact HTTP Origin/Host authority, explicit `--allow-non-loopback`, token-protected reads, disabled remote mutations, tests, and security docs | The packet's separate environment acknowledgement is not implemented; preserve the deliberate flag-plus-read-only-token design or revisit that decision |
| P1-1 Shared-service auth/RBAC | Conditional | No supported shared-service mode | Required before any multi-user or Internet-facing deployment |
| P1-2 Threat model | Implemented | `docs/threat-model.md`, `SECURITY.md`, dashboard guidance, strict-doc tests, and docs navigation | Keep residual risks synchronized as isolation and ingestion controls change |
| P1-2 Execution isolation | Backlog | Existing pinning, path, environment, and redaction controls are not a sandbox | Add process, filesystem, credential, and egress isolation before treating remote code as safely executable |
| P1-3 SBOM and attestations | Implemented locally | Release workflow and validator cover all runtime extras and attest distributions, SBOM, and five graph release assets | Verify attestations against the next real tagged release |
| P1-4 Shared deployment | Conditional | OTLP export already exists for the local product | Docker, readiness, identity, RBAC, and service operations are required only if shared-service scope is adopted |
| P2-1 Package migration | Backlog | `pyproject.toml` still ships legacy flat `py-modules` | Finish migration with compatibility shims and removal dates |
| P2-2 Coverage and typing | Backlog | CI coverage floor remains 40%; mypy still ignores missing imports | Raise the floor in measured steps and tighten core recommend/install paths |
| P2-3 Reproducible install | Partial | The no-test policy gate has a platform-specific hash lock; release artifacts and dependency closure are validated | Add and verify a maintained hash-locked default installation artifact for users |
| P2-4 Oversized modules | Backlog | Several source and test modules remain over the local contribution threshold | Split by responsibility after separate dead-code cleanup commits |
## Review Corrections
The packet is directionally strong, especially on governance, provenance, and
operability. The implementation must make these corrections:
- A declared license or arbitrary permission string is not legal evidence.
Full-body ingestion must bind immutable evidence and remain blocked when that
evidence is absent.
- The registry-enforced Design.md import path fails closed, but other catalog,
recommendation, installation, and publication paths do not yet share that
enforcement boundary. Policy intent must not be described as universal
runtime behavior.
- An SBOM generated from only the base wheel environment would be incomplete
for a product that advertises optional runtimes. The implemented release
contract therefore covers all six runtime extras and attests graph assets as
well as distributions and the SBOM.
- A local loopback monitor is not a shared service. A non-loopback opt-in does
not substitute for identity, RBAC, CSRF controls, isolation, or auditability.
- The local monitor deliberately substitutes one explicit non-loopback flag,
token-protected reads, and disabled remote mutations for the packet's
flag-plus-environment-acknowledgement proposal.
- The active GitHub ruleset CI-protects `main`, but it does not require
pull-request review or CODEOWNER approval. CODEOWNERS names only one human,
so independent review remains a human/operator blocker.
- Functional loopback benchmark evidence must not be labeled as live-provider
or production-efficiency evidence.
- Dependency automation now uses a fail-closed structural exemption in the
no-test policy, and the policy gate's own dependencies are hash locked.
Remote Dependabot and audit execution still needs final-branch evidence.
## Current Provenance Risk
The current catalog audit found a large inherited corpus with sparse license
metadata. The registry-governed Design.md import path responds by failing
closed:
- metadata-only records may remain discoverable with
`installable=false`;
- unknown-license full bodies are not approved for redistribution or install;
- imported flags never override derived provenance policy;
- historical bodies require quarantine, verified evidence, or human legal
disposition rather than guessed licenses.
Approved full-body imports now require digest-verified checked-in legal
evidence and a versioned canonical full-manifest binding. The inherited
Design.md corpus remains blocked because it lacks that complete evidence.
The source registry and `NOTICE` are controls and evidence indexes; they do not
constitute legal advice or prove ownership, compatibility, or redistribution
rights.
Other catalog, recommendation, installation, and publication paths can still
derive availability or installability without consulting this registry.
Universal provenance enforcement therefore remains a critical partial item.
## Review Batch Exit Criteria
This repository review batch is complete only when:
1. every active patch above has an independent reviewer verdict;
2. canonical QA trackers match committed code and contain no stale completion
prose;
3. local-fast, full tests, strict docs, package, security, and browser gates
pass;
4. remote required checks pass on the final branch;
5. human-owned blockers remain explicitly blocked rather than reported as
complete;
6. the final no-mistakes run passes after all product phases are finished.
Passing those gates closes this review batch; it does not by itself make ctx
enterprise ready.
## Enterprise-Ready Exit Criteria
An enterprise-ready claim additionally requires:
1. every in-scope P0 and P1 criterion above to be implemented rather than
partial, pending, or blocked;
2. independent ownership and tested pull-request and CODEOWNER review controls;
3. universal provenance enforcement across ingestion, recommendation,
installation, and publication;
4. observed scheduled security-scanner runs and verification against a real
tagged release; and
5. identity, authorization, and isolation controls before any shared-service
deployment is supported.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 472k | Updated | Claude CodeCodex | Verified | Install |
| 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: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |