pr-sentinel 2.3.0a1

Local PR review tool that orchestrates AI agents over git diffs via GitHub Copilot or Claude Code CLI.

PR Sentinel

Local pull-request review tool. Reads a git diff, runs four specialized review agents over it via a local AI CLI, and emits a structured JSON + Markdown report so you can fix issues before raising the PR.

With --pr, the review diffs the PR's own branches, checks whether the change satisfies the PR's linked Azure DevOps work item(s) (--align), and ends in an interactive HTML page where you can select findings, gaps, and alignment verdicts to push back to the PR as comment threads — all in one command.

No API keys. No hosted services. PR Sentinel shells out to a provider CLI you already have — the Claude Code CLI (default) or the GitHub Copilot CLI (--provider copilot) — and uses that tool's existing authentication. See Providers.

Reports come out as JSON, Markdown, or a self-contained HTML page with editor deep-links (--format).

How it works

git diff main...HEAD  (or PR branches when --pr is given)
        │
        ▼
[ diff_parser ] ── filters built-in noise + user skip patterns
        │             (--skip-files, .prsentinelignore)
        ▼
[ router ] ── per file, decides which agents are relevant for those file types
        │
        ▼
[ chunker ] ── packs each agent's files into ≤100k-char chunks (hybrid batching)
        │
        ▼
[ orchestrator ] ── runs (agent × chunk) tasks in a bounded thread pool
        │
        ├── Security Agent      (provider CLI + prompts/security.md)
        ├── Code Quality Agent  (provider CLI + prompts/quality.md)
        ├── Performance Agent   (provider CLI + prompts/performance.md)
        └── Testing Agent       (provider CLI + prompts/testing.md)
                │
                ▼
[ cache ] ── sha256(provider + model + prompt) → response, on disk
                │
                ▼
[ Summary Agent ] ── single provider call that dedupes/merges findings across agents
        │             (prompts/summary.md; falls back to raw findings on failure)
        │
        ▼    ← with --align (or `review-alignment`)
[ Alignment Agent ] ── one provider call per linked work item, whole diff in one prompt
        │               judges whether each acceptance criterion is satisfied
        │               (prompts/alignment.md; diff capped at 500k chars)
        ▼
[ report_generator ] ── merges code findings + gap findings + alignment verdicts
        │                  renders JSON / Markdown / HTML (--format)
        ▼
reports/report.json + reports/review-report.md (+ review-report.html)
        │
        ▼    ← with --pr (or `push-azure`)
[ push_server ] ── local server backing the interactive HTML report
                   tick findings / gaps / verdicts → push to Azure DevOps PR

File-type routing

Not every agent has something useful to say about every file. PR Sentinel skips agents that have nothing meaningful to contribute to a chunk's file types, reducing token usage with no loss in review quality.

File type	Security	Quality	Performance	Testing
*.cs (source)	✅	✅	✅	✅
*Test*.cs, *Spec.cs, *Fixture.cs	✅	✅	—	✅
*.cshtml, *.razor, *.html	✅	✅	—	—
*.js, *.ts	✅	✅	✅	✅
*.css, *.scss, *.less	—	✅	—	—
*.svg	✅	—	—	—
*.csproj, *.props, *.targets	✅	✅	—	—
*.sln	—	—	—	—
appsettings*.json, *.yml, *.yaml, *.xml, *.json	✅	✅	—	—
Dockerfile, *.bicep, *.tf	✅	✅	—	—
*.sql	✅	✅	✅	—
*.csv, *.tsv, *.resx	✅	—	—	—
*.md, *.txt, *.http	✅	—	—	—
Unknown extension	✅	✅	✅	✅

Three rules behind the table:

• Security runs on almost everything — secrets and PII appear in docs, config files, and data files.
• Performance and Testing only run on executable code.
• Unknown extensions always get all four agents — nothing is silently skipped.

If a chunk contains mixed file types (e.g. a .cs file and a .css file together), the agents for the union of both types are run. Binary files (*.png, *.dll, *.zip, etc.) are dropped entirely before routing — their diffs are unreadable.

Requirements

• Python 3.11+
• At least one supported provider CLI, installed and authenticated:
  • Claude (default) — Claude Code CLI. claude --version must work from your shell.
  • Copilot (optional, --provider copilot) — GitHub Copilot CLI. copilot --version must work, and you must have run copilot login.
• Git, if you want to review live branches (not required for --diff mode).
• An Azure DevOps Personal Access Token when using --pr, --align, push-azure, or review-alignment. See Azure DevOps PAT.

Install

From PyPI:

pip install pr-sentinel

From source (for development):

git clone https://github.com/kishor2004reddy/Pull-Request-Sentinel-CLI.git
cd Pull-Request-Sentinel-CLI
pip install -e .[dev]

Verify:

pr-sentinel --help
pr-sentinel agents

Quickstart

Review your current branch against main:

pr-sentinel review --base main

Or review a saved diff file (works without a git repo):

git diff main...HEAD > my.diff
pr-sentinel review --diff my.diff --out reports

Open reports/review-report.html.

Review a PR, push findings to it, and also check requirement coverage:

# Diffs the PR's own branches, asks for a PAT interactively, runs the code
# review + alignment agent, then opens a push page in your browser.
pr-sentinel review --pr 124 --align

Commands

pr-sentinel review

Review a diff and write a structured report. With --pr, the command also opens an interactive push page.

Flag	Default	Description
--base	main	Branch to diff against. Runs git diff <base>...<head>.
--head	HEAD	Source branch/ref to review. Use with --base to diff arbitrary refs without checking them out.
--diff PATH	—	Review a saved diff file instead of running git. Mutually exclusive with --staged.
--staged	off	Review staged changes (git diff --cached).
--fetch	off	Fetch both sides of the diff from the remote before diffing. Rewrites base → {remote}/{base} and head → {remote}/{current_branch} so the review matches exactly what Azure DevOps shows. Enabled automatically when --pr derives branches.
--remote	origin	Git remote to fetch from and to use for Azure DevOps remote detection. Use when your Azure remote is not named origin.
--repo PATH	cwd	Path to the git repository to review. Ignored when --diff is used.
--agents	security,quality,performance,testing	Comma-separated agents to run.
--out	./reports	Output directory.
--format	both	Report format(s): json, markdown, html, both (json+html, default), or all (json+html+markdown).
--max-file-size	20000	Per-file diff size cap (chars). Larger files get truncated with a marker.
--chunk-budget	100000	Max combined diff size per provider call before chunking kicks in. Does not affect the alignment agent (see ALIGNMENT_DIFF_BUDGET).
--provider	claude	AI CLI to run the agents through. claude shells out to claude -p; copilot shells out to the GitHub Copilot CLI. See Providers.
--model	provider default	Model to use, forwarded verbatim to the selected provider. claude: shortcuts sonnet, opus, haiku, or a full ID like claude-opus-4-8, claude-sonnet-4-6 (default sonnet). copilot: a Copilot model ID such as claude-sonnet-4.6, gpt-5 (default claude-sonnet-4.6).
--max-parallel	12	Max concurrent provider calls across all (agent, chunk) pairs.
--timeout	600	Per-call timeout in seconds for each provider subprocess.
--no-cache	off	Bypass the response cache for this run. Successful responses are still written to the cache.
--skip-files	—	Comma-separated glob patterns to skip on top of built-in noise filters. Combines with .prsentinelignore if present.
--pr ID	—	Azure DevOps pull request ID. Auto-derives base/head branches from the PR, fetches them, runs the review against the PR's own diff, and ends with an interactive push page. Requires a PAT (prompted interactively if not set in the environment).
--align	off	After the code review, run the Alignment Agent to check whether the change satisfies the PR's linked work item(s). Requires --pr. The combined report shows code findings, alignment verdicts, and gaps all in one page.
--org	auto	Azure DevOps organization (overrides remote detection; used with --pr).
--project	auto	Azure DevOps project (overrides remote detection; used with --pr).
--azure-repo	auto	Azure DevOps repository name (overrides remote detection; used with --pr). Distinct from --repo, which is the local path.
--port	0	Local push-server port (with --pr). 0 picks a free port.
--no-browser	off	Don't auto-open the combined report in a browser (with --pr).

pr-sentinel review-alignment

Standalone requirement-coverage review. Fetches the work item(s) linked to a PR (or a specific work item), diffs the branches, and runs the Alignment Agent to judge whether each acceptance criterion is satisfied.

# Review work items linked to PR #124 against the current branch
pr-sentinel review-alignment --pr 124

# Review a specific work item against an explicit diff range
pr-sentinel review-alignment --work-item 456 --base main --head feature/csv-export

# Use with a non-default remote
pr-sentinel review-alignment --pr 124 --remote azure

Flag	Default	Description
--pr ID	—	Azure DevOps PR whose linked work items are reviewed.
--work-item ID	—	Review a specific work item directly (without a PR). Mutually exclusive with --pr.
--base	main	Base branch to diff against.
--head	HEAD	Source branch/ref.
--fetch	off	Fetch branches from the remote before diffing. Enabled automatically when --pr derives branches.
--remote	origin	Git remote for Azure detection and --fetch.
--repo-dir PATH	cwd	Repository to diff and whose remote is used for detection.
--out	./reports	Output directory for the alignment report.
--provider	claude	AI CLI.
--model	provider default	Model forwarded to the provider.
--timeout	600	Per-call timeout in seconds.
--no-cache	off	Bypass the response cache.
--org / --project / --repo	auto	Override org/project/repo (otherwise parsed from the origin remote).

Writes reports/alignment-report.json and reports/alignment-report.html. To post the gaps and verdicts to a PR:

pr-sentinel push-azure --pr 124 --report reports/alignment-report.json

pr-sentinel agents

Lists available agents and their implementation status.

pr-sentinel push-azure

Push selected findings, gaps, and/or alignment verdicts from a prior review to an Azure DevOps pull request as PR-level comment threads. You pick which items to push interactively in the HTML report.

# 1. Review (HTML + JSON are written by default)
pr-sentinel review --base main

# 2. Push selected findings to PR #124 (PAT is prompted interactively)
pr-sentinel push-azure --pr 124

Or push the output of a review-alignment run:

pr-sentinel push-azure --pr 124 --report reports/alignment-report.json

This starts a small local server on 127.0.0.1, opens the report in your browser, and adds checkboxes plus a "Push selected to PR" button. Tick the items you want, click push, and each becomes a comment thread on the PR. Press Ctrl+C in the terminal when done.

Flag	Default	Description
--pr	required	Azure DevOps pull request ID to comment on.
--report PATH	reports/report.json	Report produced by a prior review or review-alignment run.
--org / --project / --repo	auto-detected	Override the org/project/repo (otherwise parsed from the origin remote).
--remote	origin	Git remote whose URL is parsed for org/project/repo detection.
--repo-dir PATH	cwd	Repository whose origin remote is parsed for the org/project/repo.
--port	0	Local server port (0 picks a free one).
--no-browser	off	Don't auto-open the browser.

Notes:

• The PAT never reaches the browser. It is read from AZURE_DEVOPS_PAT (or SYSTEM_ACCESSTOKEN in Azure Pipelines), or prompted interactively if neither is set. It is used only by the local server for authenticated REST calls, never stored to disk, and never pushed into os.environ (which would leak it into provider subprocesses).
• Idempotent for findings. Each finding thread is tagged with the finding's stable id, so re-pushing the same finding is skipped rather than duplicated.
• Upsert for alignment verdicts. Alignment verdict threads are tagged separately. Re-pushing an already-posted verdict patches the existing comment, so a second push always refreshes it on the PR.
• Already-pushed detection. On page load, the server queries Azure DevOps live and marks items that are already on the PR. Soft-deleted comment threads (deleted in Azure but not yet re-pushed) are excluded from the "already pushed" count.
• Auto-detection. The org/project/repo come from your origin remote (HTTPS, visualstudio.com, or SSH forms). Pass --org/--project/--repo to override or if there's no Azure DevOps remote.

pr-sentinel config

Set personal defaults for review flags so you don't have to repeat them on every run. Settings are stored in ~/.pr-sentinel/config.toml and apply to the review command. Explicit CLI flags always take precedence over config defaults.

# Set a default provider
pr-sentinel config set provider copilot

# Run only security and quality agents by default
pr-sentinel config set agents security,quality

# Change the default base branch
pr-sentinel config set base develop

# Reduce parallelism (useful if you hit rate limits)
pr-sentinel config set max-parallel 6

# See all keys, your overrides, and built-in defaults
pr-sentinel config list

# Revert a single key to its built-in default
pr-sentinel config unset provider

# Clear every override at once
pr-sentinel config reset

Subcommand	Description
config set <key> <value>	Set a default. Key names match flag names without leading dashes (max-parallel or max_parallel both work). Value is validated immediately.
config unset <key>	Remove a single default, reverting that key to its built-in value.
config list	Show all configurable keys, your current overrides, and the built-in defaults.
config reset	Remove all overrides (prompts for confirmation). Deletes the config file.

Configurable keys:

Key	Built-in default	Accepts
provider	copilot	claude, copilot
model	provider default	any model string
agents	security,quality,performance,testing	comma-separated subset
base	main	any branch/ref
remote	origin	any remote name
fetch	false	true, false
format	both	json, markdown, html, both, all
out	./reports	any directory path
max_parallel	12	positive integer
timeout	600	positive integer (seconds)
max_file_size	20000	positive integer (chars)
chunk_budget	100000	positive integer (chars)

Precedence: built-in defaults → ~/.pr-sentinel/config.toml → CLI flags

pr-sentinel cache

Inspect and manage the on-disk response cache.

Subcommand	Description
cache size	Show cache location, entry count, and disk usage.
cache clear	Wipe the entire cache (prompts for confirmation).
cache prune --older-than 30d	Delete entries older than the given age. Supports s/m/h/d suffixes. Add --dry-run to preview.

The cache lives at ~/.pr-sentinel/cache/ by default. Override with the PR_SENTINEL_CACHE_DIR environment variable. Keys are sha256(provider + model + prompt), so changing the provider, the model, or any prompt content invalidates the entry automatically (and a Claude run never collides with a Copilot run that happens to use the same model name).

Auto-pruning. Every pr-sentinel review run silently drops cache entries older than 90 days before doing any work. No flag, no output — it just keeps the cache from growing unbounded over time. The threshold is set via AUTO_PRUNE_AGE_DAYS in config.py. Use cache prune --older-than ... for manual prunes at a different age.

Azure DevOps PAT

Commands that talk to Azure DevOps (review --pr, review --align, review-alignment, push-azure) need a Personal Access Token.

Interactive prompt (recommended for local use). If neither AZURE_DEVOPS_PAT nor SYSTEM_ACCESSTOKEN is set and stdin is a terminal, PR Sentinel prompts for the PAT with hidden input. The value lives only in the current process's memory for the run — it is never stored to disk and is deliberately not pushed into os.environ (which would leak it into the provider CLI subprocesses that are spawned during the review).

Environment variable (recommended for CI). Set AZURE_DEVOPS_PAT before running. If you're inside an Azure Pipelines job, SYSTEM_ACCESSTOKEN is also recognized.

# PowerShell
$env:AZURE_DEVOPS_PAT = "<your-pat>"

# bash
export AZURE_DEVOPS_PAT=<your-pat>

Required scopes by command:

Command	Required PAT scopes
review --pr	Code: read & write
review --pr --align	Code: read & write, Work Items: read
review-alignment	Work Items: read, Code: read
push-azure	Code: read & write

Alignment Agent

The Alignment Agent is a separate, holistic review pass that checks whether the PR's code changes actually satisfy each work item's acceptance criteria. Unlike the four code-review agents (which chunk large diffs), the Alignment Agent receives the entire diff in a single provider call per work item, so it can reason about cross-file interactions that would be lost in chunks.

How it works:

1. Fetches the work item(s) linked to the PR from Azure DevOps.
2. Formats the work item — title, description, acceptance criteria, repro steps (for bugs).
3. Sends one prompt per work item: the formatted work item + the complete diff.
4. Returns a verdict (Satisfied / Partial / Not satisfied), a confidence rating, a summary, per-criterion trace (Met / Partial / Not met / Unverifiable), and a list of gap findings (pushable as PR comments).

Diff budget. The alignment agent uses ALIGNMENT_DIFF_BUDGET = 500,000 chars (about 150k–165k tokens), separate from the routed agents' DEFAULT_CHUNK_BUDGET = 100,000 chars. This covers virtually every real PR in one call while staying well clear of the model's context window. Diffs above the budget are truncated with a visible warning.

Triggering alignment:

• review --pr <id> --align — runs alignment after the code review; shows everything in one combined HTML report.
• review-alignment --pr <id> — standalone alignment pass; writes alignment-report.json and alignment-report.html.

Providers

PR Sentinel doesn't talk to any AI service directly — it shells out to a provider CLI you already have installed and authenticated. Pick one with --provider:

Provider	CLI invoked	Prompt delivery	Default model
claude (default)	claude --model <m> -p	stdin	sonnet
copilot	copilot --no-color [--model <m>]	stdin	claude-sonnet-4.6

Notes:

• Models are provider-specific. --model is forwarded verbatim to whichever provider you select; there's no translation between namespaces. sonnet means something to Claude, gpt-5 means something to Copilot. If you pass a model the provider doesn't offer, that CLI reports the error.
• Discovering Copilot models. Available models depend on your GitHub Copilot plan and can't be listed non-interactively. To see what your account can use, run copilot and type /model. PR Sentinel defaults to claude-sonnet-4.6 for Copilot; override with --model if your plan doesn't include it.
• Copilot runs read-only. PR Sentinel invokes Copilot without --allow-all-tools. The full diff is embedded in the prompt and the agent only returns a JSON verdict, so Copilot never needs to run shell commands or edit files in your repo.
• Authentication is the provider's. No API keys live in PR Sentinel — Claude uses your Claude Code login, Copilot uses your copilot login session.

# Run the review through Claude (default)
pr-sentinel review --base main

# Run the review through GitHub Copilot instead
pr-sentinel review --base main --provider copilot

# Pick a specific Copilot model
pr-sentinel review --diff my.diff --provider copilot --model claude-sonnet-4.5

Skipping files

PR Sentinel always filters built-in noise (lock files, build/dist dirs, minified, generated). To skip additional files — large fixtures, vendored code, anything that wastes tokens without adding review value — use either of these:

Per-run flag, good for one-offs:

pr-sentinel review --skip-files "vendor/**,fixtures/*.json,*.snap"

Repo-wide ignore file, good for project defaults — drop a .prsentinelignore at the repo root:

# generated docs
docs/api/**

# huge test fixtures
tests/fixtures/large/**

# vendored deps
vendor/**

Syntax: one glob per line, # for comments, blank lines ignored. Patterns use fnmatch (same as the built-in noise filter), matched against both the full path and the basename. The two sources combine — flag patterns are appended to the ignore file's patterns. Skipped files are listed in the "Skipped noise file(s)" panel at the top of the run output.

Examples

Review current branch vs main:

pr-sentinel review --base main

Run only the security agent:

pr-sentinel review --base main --agents security

Review what's currently staged:

pr-sentinel review --staged

Review a feature branch without checking it out:

pr-sentinel review --base main --head feature/new-auth

Use a stronger model for higher-stakes reviews:

pr-sentinel review --base main --model opus

Also emit a Markdown report (HTML + JSON are the default):

pr-sentinel review --base main --format all

Force a fresh run, ignoring cached responses:

pr-sentinel review --base main --no-cache

Skip extra files for this run only:

pr-sentinel review --base main --skip-files "vendor/**,*.snap"

Review a PR and push findings — interactive, one command:

# Auto-derives branches, prompts for PAT, opens push page when done.
pr-sentinel review --pr 124

Review + alignment in one pass — push findings, gaps, and verdicts:

pr-sentinel review --pr 124 --align

Standalone alignment review for a PR:

pr-sentinel review-alignment --pr 124
# Then push the alignment output:
pr-sentinel push-azure --pr 124 --report reports/alignment-report.json

Alignment for a specific work item (no PR):

pr-sentinel review-alignment --work-item 456 --base main --head feature/csv-export

Alignment via non-default remote:

pr-sentinel review --pr 124 --align --remote azure

Prune cache entries older than a week:

pr-sentinel cache prune --older-than 7d

Set personal defaults so you never have to repeat them:

pr-sentinel config set provider copilot
pr-sentinel config set agents security,quality
pr-sentinel config set base develop
pr-sentinel config list

Report structure

PR Sentinel can emit several report formats (pick with --format), all built from the same underlying report object:

• report.json — the structured report, suitable for piping into other tools.
• review-report.md — the human-readable Markdown report.
• review-report.html — a single self-contained HTML page (inline CSS/JS, no external assets) with severity badges and vscode:// editor deep-links straight to each finding's file.
• alignment-report.json / alignment-report.html — written by review-alignment; contain the alignment verdict sections and gap findings.

--format both (the default) writes JSON + HTML; --format all adds Markdown. Every run also drops the raw diff it reviewed at reports/source.diff. At the end of each run the CLI prints a Run Stats panel — total time, provider calls, cache hit rate, and (when the provider reports them) tokens, cost, and Copilot premium requests.

The review command automatically selects the HTML renderer based on report content: combined when both code findings and alignment are present, alignment-only when only alignment ran, or a plain code review page otherwise.

Code review HTML report

A self-contained HTML page with:

• Overview tab — risk level, merge verdict, PR section (when --pr was used: PR number linked to Azure DevOps, repo, base branch, source branch), agent summary table.
• Code Review tab — all findings with severity badges, file paths, line hints, reasoning, and vscode:// deep-links.
• Interactive push toolbar — checkboxes per finding/gap/verdict, a "Push selected to PR" button, and real-time "already pushed" detection on page load.

Combined HTML report (review --pr --align)

When both code review findings and alignment verdicts are present, a single combined page is rendered:

• Overview tab — PR info, risk level, agent summary.
• Code Review tab — code findings from the four routed agents.
• Alignment tab — per-work-item verdict badge, acceptance-criterion traceability matrix, coverage bar, and gap findings.
• Gaps tab — gap findings surfaced by the Alignment Agent, pushable as PR threads.
• One push toolbar that handles all three item types: code findings, gap findings, and alignment verdicts.

Alignment HTML report (review-alignment)

Stand-alone alignment report with:

• Per-work-item verdict badge (Satisfied / Partial / Not satisfied) and confidence.
• Acceptance-criterion traceability matrix with Met / Partial / Not met / Unverifiable status and evidence snippets.
• Coverage bar.
• Gap findings section.
• Push toolbar (gaps and verdicts are pushed via push-azure --report reports/alignment-report.json).

Markdown report sections

The markdown report always emits these five sections in this fixed order, regardless of findings:

1. Summary — risk level, source, branch, timestamp, agents, finding counts
2. Merge Verdict — deterministic verdict driven by risk level (not a separate Claude call)
3. Key Findings — top blocking issues (High + Medium)
4. Key Recommendations — deduplicated fixes
5. All Findings — per-agent summary table + full per-issue detail

JSON report

The JSON report contains the same data in a single structured object suitable for piping into other tools. Key fields:

• riskLevel — High / Medium / Low / None / Unknown (see table below)
• coverageComplete — false if any agent failed during the run
• agentsExecuted, failedAgents — who ran and who didn't
• findings — flat list, sorted by severity then file; each finding has a stable id for push tracking
• rawFindingCount — only present when the Summary Agent ran; how many findings existed before dedup
• alignment — list of per-work-item verdict sections (present when --align was used or review-alignment was run)
• pr — PR context object (present when --pr was used): id, org, project, repo, title, baseBranch, sourceBranch

Risk levels

Level	Trigger
High	Any High-severity finding, or 5+ Medium-severity findings
Medium	One or more Medium-severity findings (and fewer than 5)
Low	Only Low-severity findings
None	No findings across all executed agents (full coverage)
Unknown	All executed agents failed, or some agents failed and the rest found nothing (can't call it clean with incomplete coverage)

Architecture

src/pr_sentinel/
├── cli.py                  # Click entrypoint — wires the pipeline together
│                             (review, review-alignment, push-azure, agents, cache, config)
├── ui.py                   # Rich panel/table builders (pure: data → renderable)
├── config.py               # tunable defaults + shared constants (single source of truth)
│                             ALIGNMENT_DIFF_BUDGET (500k chars, separate from chunk budget)
│                             load_user_config / save_user_config → ~/.pr-sentinel/config.toml
├── runstats.py             # thread-safe per-run metrics (calls, tokens, cost, time)
├── diff/
│   ├── git_diff.py         # git rev-parse, git diff, --staged, fetch, get_current_branch
│   ├── diff_parser.py      # per-file splitting + noise filter (+ --skip-files / .prsentinelignore) + truncation
│   └── chunker.py          # greedy packer to keep prompts under chunk-budget
├── providers/
│   ├── __init__.py         # provider dispatch: name -> runner module (get_runner)
│   ├── common.py           # shared JSON extraction + cache + 1-retry logic + runstats
│   ├── claude.py           # subprocess(claude -p), prompt on stdin
│   └── copilot.py          # subprocess(copilot --no-color), prompt on stdin
├── orchestrator.py         # parallel (agent, chunk) execution via ThreadPoolExecutor
├── router.py               # file-type routing table — decides which agents run per chunk
├── cache.py                # sha256-keyed disk cache + 90-day auto-prune
├── push_server.py          # local server backing the HTML report's "Push" button
│                             /pushed endpoint: live-queries Azure for already-pushed finding ids
│                             and alignment verdict ids (handles Azure soft-delete)
├── integrations/
│   └── azure_devops.py     # remote parsing + Azure DevOps PR client
│                             list_thread_finding_ids, list_alignment_work_item_ids,
│                             upsert_alignment_comment, get_pr_work_items, get_work_items
├── report_generator/
│   ├── __init__.py         # build_report + JSON/Markdown/HTML writers + shared helpers
│   ├── markdown.py         # Markdown renderer
│   └── html.py             # self-contained HTML renderer (tabbed, editor deep-links)
│                             _render_html (code review), _render_alignment_html,
│                             _render_combined_html (review + alignment in one page)
├── agents/
│   ├── base.py             # BaseAgent: load prompt, chunk, call, validate
│   ├── security_agent.py
│   ├── quality_agent.py
│   ├── performance_agent.py
│   ├── testing_agent.py
│   ├── summary_agent.py    # post-processes findings: dedupe/merge across agents
│   └── alignment_agent.py  # holistic per-work-item verdict (whole diff in one call)
└── prompts/
    ├── security.md
    ├── quality.md
    ├── performance.md
    ├── testing.md
    ├── summary.md
    └── alignment.md

tests/
├── test_diff_parser.py
├── test_chunker.py
├── test_providers.py
├── test_summary_agent.py
├── test_report_generator.py
├── test_azure_devops.py
├── test_push_server.py
└── test_cli.py

Running tests

pip install -e .[dev]
pytest -q

Troubleshooting

claude CLI not found on PATH — Claude Code is the default provider. Install Claude Code and confirm claude --version works in the same shell where you run pr-sentinel.

copilot CLI not found on PATH — only relevant when using --provider copilot. Install the GitHub Copilot CLI (npm install -g @github/copilot-cli) and confirm copilot --version works in the same shell.

copilot returned non-JSON output after retry — the provider CLI occasionally returns prose instead of JSON. The runner retries once; if it still fails, that chunk's findings are dropped and the agent is marked failed for the run. Re-running usually succeeds.

Slow runs — large diffs trigger chunking. Each chunk is one provider call per agent. Reduce scope with --agents security if you only want one perspective, or with --max-file-size to truncate huge files. The cache amortizes repeat runs against the same diff.

Lock files / minified files showing up — they shouldn't. If they do, add the pattern to NOISE_PATTERNS in config.py, or skip on a per-project basis with .prsentinelignore (see Skipping files).

Want to skip a non-noise file (huge fixture, vendored dep, generated snapshot)? Use --skip-files "pat1,pat2" for a one-off, or commit a .prsentinelignore for project-wide defaults.

No Azure DevOps PAT found — run in a terminal (not piped) to get the interactive PAT prompt, or set AZURE_DEVOPS_PAT in your environment before running. In Azure Pipelines, set SYSTEM_ACCESSTOKEN. See Azure DevOps PAT for required scopes.

No linked work items found for PR #… — the PR must have Azure DevOps work items linked to it (via the "Related work items" section in the PR). If none are linked, --align will still run the code review but skip the alignment pass.

Alignment verdict shows "pushed" but comment is not visible in Azure — this happens when a PR comment thread was soft-deleted in Azure (Azure keeps the thread with the comment marked as deleted). PR Sentinel detects this and re-posts the verdict as a fresh thread when you push again. If a verdict still isn't appearing, try re-pushing via the HTML report.

Gap id changed between runs (item shows as pushable again after being pushed) — gap finding ids are derived from the finding's content (agent + file + issue text). If the model rephrases the issue slightly between runs, the id changes. This is expected behavior; push the new gap if it still represents a real issue.

Notes & limitations

• The Performance Agent only sees the diff. It can flag pattern-level issues (N+1 queries, sync-over-async) but cannot reason about runtime behavior or system load.
• lineHint is approximate — unified diffs have hunk headers, not absolute line numbers. The prompt asks the model for a description of the location rather than a hallucinated number.
• Agents cannot read other files in the repo. Review depth is limited to what's visible in the diff itself.
• If any one chunk fails for an agent (timeout, exit code, unparseable JSON after retry), that agent is marked failed for the run and its partial findings are discarded. Other agents continue.
• The Summary Agent makes one additional provider call per run to deduplicate and merge findings across the four review agents. If it fails (timeout, bad JSON), the report falls back to the raw findings — nothing is lost.
• The Alignment Agent makes one provider call per linked work item, each carrying the whole diff (up to 500k chars). For PRs with many linked work items, this runs sequentially. Very large diffs are truncated at 500k chars with a visible warning; the verdict is then partial.

License

MIT