cureview 0.1.8

CURe PR review sandboxes powered by ChunkHound and configurable LLM presets

CURe

CURe ("Code Under Review") is a CLI for running pull request reviews inside isolated sandboxes, with ChunkHound-backed code search/research and a configurable review agent on top.

It is for two audiences:

• human operators who want a repeatable way to hand PR review work to agents without letting them mutate the source checkout
• agentic sessions that need a clear, reusable bootstrap contract for fresh or partially configured environments

If you are an agent, or you want to install CURe as a reusable skill, start with SKILL.md.

Quick Links

• Why CURe
• Install And First Review
• Example Flows
• Agent And Setup Notes
• Core Commands
• Secondary Standalone Install
• Advanced / Pre-Provisioned Environments
• Minimal Config
• Jira CLI
• Tests

Why CURe

Use CURe when you want to:

• review a GitHub PR from a disposable sandbox instead of the working repo
• standardize how humans and agents start, observe, resume, and clean review runs
• give an agent a single documented path from fresh install or existing local setup to "review in progress"

CURe is different from an ad-hoc manual agent review because the project checkout stays untouched, the review state stays on disk, and the workflow is resumable instead of prompt-by-prompt improvisation.

CURe is not for:

• ad-hoc in-place repo review where the agent should work directly in the project checkout
• environments that cannot install tools or authenticate the required external systems

Install And First Review

For a human operator, the default kickoff is one sentence:

use <CURE_REPO_URL> to review <PR_URL>

Example:

use https://github.com/grzegorznowak/CURe to review https://github.com/chunkhound/chunkhound/pull/220

That sentence is the kickoff contract, not a promise that every sandbox can finish setup unattended. It tells the agent to bootstrap CURe, inspect any existing local setup, and keep the review inside cure rather than falling back to a manual review.

The operator should not need to provide a local checkout path, a config path, or a handwritten bootstrap recipe in the primary flow. If the sandbox still lacks required prerequisites such as network access or local secrets, the agent should surface the exact blocker and ask the operator for help instead of pretending CURe always self-bootstraps from zero state.

If the operator asked to use CURe, the agent should bootstrap CURe and run cure. It should not do a manual review outside CURe.

Persistent human install:

uv tool install cureview
cure init
cure install
cure doctor --pr-url <PR_URL> --json
cure pr <PR_URL> --if-reviewed new

cure install reuses an existing chunkhound already on PATH by default. Pass --chunkhound-source release or --chunkhound-source git-main only when you want CURe to install or replace that binary explicitly.

Ephemeral agent run path:

uvx --from cureview cure init
uvx --from cureview cure install
uvx --from cureview cure doctor --pr-url <PR_URL> --json
uvx --from cureview cure pr <PR_URL> --if-reviewed new

Keep the README focused on the landing page and first success. For the full agent bootstrap contract, including local setup inspection rules and operator handoff wording, use SKILL.md.

Example Flows

Example 1: clean public package install to first review

This is the primary public path and matches the package prove-out used for the first successful public release:

uv tool install cureview
cure init
cure install
cure doctor --pr-url https://github.com/chunkhound/chunkhound/pull/220 --json
cure pr https://github.com/chunkhound/chunkhound/pull/220 --if-reviewed new

The v0.1.2 public release prove-out verified that uv tool install cureview, cure init, cure install, and cure doctor --pr-url https://github.com/chunkhound/chunkhound/pull/220 --json all worked in a clean temp-home install, and that the installed tool exposed cure without the deprecated reviewflow CLI.

Example 2: ephemeral agent bootstrap from the one-sentence kickoff

Use the canonical agent run surface when the review happens inside a disposable sandbox or agent session:

tmp_root="$(mktemp -d)"
export XDG_CONFIG_HOME="$tmp_root/config"
export XDG_STATE_HOME="$tmp_root/state"
export XDG_CACHE_HOME="$tmp_root/cache"

uvx --from cureview cure init
uvx --from cureview cure install
uvx --from cureview cure doctor --pr-url <PR_URL> --json
uvx --from cureview cure pr <PR_URL> --if-reviewed new

If CURe is already partially configured, inspect the active local setup before creating a fresh one:

- the active `cure.toml`
- the JSON file resolved from `[chunkhound].base_config_path`
- repo-root `chunkhound.json` and `.chunkhound.json` as ask-first ChunkHound setup hints

On a TTY, cure init now acts as a setup wizard. It can keep the current configured base config, adopt a repo-root chunkhound.json / .chunkhound.json, accept an absolute custom base-config path, or generate the default CURe-managed base config. If chunkhound is still missing on PATH, the wizard can also offer to run cure install before returning to the original command.

Example 3: what a finished review produces

A normal review run leaves behind resumable session state plus a review artifact with stable headings:

<session_dir>/
  meta.json
  review.md

**Summary**: ...
## Business / Product Assessment
### In Scope Issues
## Technical Assessment
### In Scope Issues

Review output uses two independent lenses:

• Business / Product Assessment uses product/ticket scope.
• Technical Assessment uses implementation scope.

Agent And Setup Notes

Ensure git, curl, and ca-certificates are present before bootstrap. Install uv if it is missing.

Use cure doctor --pr-url <PR_URL> --json as the source of truth for inspect-first setup. Its repo_local_chunkhound payload plus the repo-local-chunkhound check and executor-network advisory check surface the same setup hints in machine-readable and text forms.

If repo-local ChunkHound config exists, summarize what it contains and ask the operator whether it should be reused. Do not silently adopt it in this public contract.

Commands that actually require ChunkHound bootstrap now fail or repair earlier instead of surfacing late Missing required [chunkhound] section. errors. On a TTY, cure pr, cure resume, cure followup, cure cache prime, and cure interactive can enter the same setup wizard before side effects. On non-TTY runs, those commands fail fast and point back to cure init plus cure doctor.

On an interactive cure pr cold start with no existing CURe-managed base cache for the selected baseline, CURe may also ask whether you already have a matching ChunkHound workspace/config for that exact repo. If validation passes, CURe hot-starts the managed base cache from that workspace before running the normal top-up index. Non-TTY runs skip this prompt and build the baseline cache normally.

That indexed ChunkHound-backed path is the default and recommended public review workflow.

cure doctor --pr-url <PR_URL> --json
cure pr <PR_URL> --if-reviewed new
cure resume <session_id|PR_URL>

Once the first run is active, continue the same indexed session with cure resume <session_id|PR_URL>.

cure pr --no-index remains available only as an advanced opt-out for custom prompt flows that intentionally skip the built-in ChunkHound-backed prompts. It is not the normal or recommended path.

Built-in Codex review runs use a staged CURe-managed ChunkHound helper rather than native agent MCP wiring. CURe exports that helper through CURE_CHUNKHOUND_HELPER; the built-in prompt/proof contract is successful "$CURE_CHUNKHOUND_HELPER" search ... and "$CURE_CHUNKHOUND_HELPER" research ... execution with JSON output, and helper research satisfies the code_research requirement. Plain chunkhound search, chunkhound research, and chunkhound mcp shell usage are not the built-in Codex contract. Historical sessions may still report legacy mcp_tool_call evidence.

Helper-backed Codex runs also export PYTHONSAFEPATH=1 so a ChunkHound daemon started while reviewing the chunkhound repo does not import the checked-out repo package by accident. If helper preflight times out, inspect the persisted helper path plus daemon lock/log/runtime metadata in session status or meta.json before retrying.

Codex and Claude executor paths need internet / network access to obtain code-under-review context. In constrained agent sandboxes, treat that as an operator-visible prerequisite and ask for help instead of pretending CURe can always self-bootstrap from zero state. When cure doctor resolves Codex or Claude, look for the executor-network advisory check instead of claiming the sandbox already proved that prerequisite.

Need the full bootstrap contract for agent sessions or existing local setups? Use SKILL.md.

Core Commands

Recommended indexed review loop:

cure doctor --pr-url <PR_URL> --json
cure pr <PR_URL> --if-reviewed new
cure resume <session_id|PR_URL>

Initialize non-secret bootstrap files:

cure init

When stdin/stderr are a real terminal, cure init is interactive and acts as the guided bootstrap/repair entry point.

Start a fresh review:

cure pr <PR_URL> --if-reviewed new

Check status:

cure status <session_id|PR_URL> --json

Watch a run:

cure watch <session_id|PR_URL>

Resume a session:

cure resume <session_id|PR_URL>

Synthesize a final review:

cure zip <PR_URL>

Clean up old sessions:

cure clean closed --json
cure clean <session_id>

Show the machine-readable command catalog:

cure commands --json

Secondary Standalone Install

The public package remains the default and recommended path:

uv tool install cureview

Use the standalone GitHub Release assets only when the package path is unavailable or inconvenient. The current secondary targets are:

• Linux x86_64
• macOS x86_64
• macOS arm64

Install the latest standalone release:

curl -fsSL https://raw.githubusercontent.com/grzegorznowak/CURe/main/install-cure.sh | sh

Pin a specific standalone release:

curl -fsSL https://raw.githubusercontent.com/grzegorznowak/CURe/main/install-cure.sh | sh -s -- --version v0.1.8

The installer downloads the matching release asset into ~/.local/bin/cure. After that, the bootstrap/readiness flow is unchanged:

cure init
cure install
cure doctor --pr-url <PR_URL> --json
cure pr <PR_URL> --if-reviewed new

If your platform is not covered by the standalone assets, fall back to the package path instead of inventing a separate bootstrap recipe.

Advanced / Pre-Provisioned Environments

Persistent human install should use the public package:

uv tool install cureview
cure init
cure install
cure doctor --pr-url <PR_URL> --json
cure pr <PR_URL> --if-reviewed new

Teams that already manage a local CURe checkout can keep using that as a secondary local-development flow:

• keep CURe in a stable local path
• refresh it with git -C <CURE_SOURCE> pull --ff-only
• install it with uv tool install /path/to/cure
• keep any project-specific wrappers or config beside that checkout

If local CURe config already exists, inspect it before overwriting it:

• check the active cure.toml
• inspect the JSON resolved from [chunkhound].base_config_path
• treat repo-root chunkhound.json or .chunkhound.json as setup hints to discuss with the operator, not inputs to silently adopt

Prefer cure doctor --pr-url <PR_URL> --json as the readiness summary after that inspection. It now reports repo_local_chunkhound plus the repo-local-chunkhound and executor-network checks so the operator does not need to infer those details from raw files alone.

Those details are secondary. The primary operator contract stays use <CURE_REPO_URL> to review <PR_URL>.

For the full bootstrap contract around fresh installs, existing local setup, and operator handoff, use SKILL.md.

Minimal Config

Default config path:

~/.config/cure/cure.toml

By default cure init also writes:

~/.config/cure/chunkhound-base.json

If you need a disposable or non-default layout, set XDG_CONFIG_HOME, XDG_STATE_HOME, and XDG_CACHE_HOME before cure init, or pass --config, --sandbox-root, and --cache-root directly to cure init.

Minimal config written by cure init:

[paths]
sandbox_root = "/absolute/path/to/sandboxes"
cache_root = "/absolute/path/to/cache"

[review_intelligence]
[[review_intelligence.sources]]
name = "github"
mode = "auto"

[[review_intelligence.sources]]
name = "jira"
mode = "when-referenced"

[chunkhound]
base_config_path = "/absolute/path/to/chunkhound-base.json"

[multipass]
# strict = fail closed on invalid grounding
# warn   = record findings and continue
# off    = skip grounding validation
grounding_mode = "strict"
step_workers = 4

# Optional multipass-only reasoning-effort overrides.
# If unset:
# - plan and synth keep using the generic resolved reasoning effort
#   (the built-in `codex-cli` preset defaults that generic effort to `xhigh`)
# - step defaults to `medium` for built-in big multipass reviews
plan_reasoning_effort = "xhigh"
step_reasoning_effort = "medium"
synth_reasoning_effort = "xhigh"

When strict multipass grounding fails, CURe keeps the invalid artifact on disk and writes the validation details to work/grounding_report.json inside the session. Inspect the persisted state with cure status <session_id|PR_URL> --json or cure watch <session_id|PR_URL>, then rerun the same session with cure resume <session_id> or the narrower cure resume <session_id> --from steps / cure resume <session_id> --from synth. If you want fail-open behavior for future runs, set [multipass].grounding_mode = "warn".

If an embedding key is already present in the environment, cure init adds the matching embedding block and continues. If VOYAGE_API_KEY already exists, cure init writes the Voyage embedding model into the active ChunkHound base config and continues. Otherwise, if OPENAI_API_KEY already exists, cure init writes the OpenAI embedding model into the active ChunkHound base config and continues.

If no supported key is present, the agent should stop with the exact local config path, the minimal snippet to add, the required env var name, and the rerun command instead of improvising a manual review.

The structured review_intelligence source registry now feeds prompt guidance, session metadata, and cure doctor --json capability summaries. Only mode = "required" sources are preflighted before review generation; optional sources stay lazy and surface as available, unavailable, or unknown based on the runtime facts CURe already has.

Jira CLI

Use this only when the workflow actually needs Jira context. Normal public GitHub PR review flows do not require Jira.

For tenant setup, auth, jira init, JIRA_CONFIG_FILE, common queries, and troubleshooting, use the dedicated Jira reference.

If Jira context is required in a CURe session, keep auth local, prefer ~/.netrc for api.atlassian.com or a short-lived JIRA_API_TOKEN, and point CURe at a non-default Jira CLI config with JIRA_CONFIG_FILE when needed.

Tests

Fast local check:

./selftest.sh