Changelog

Fixes for every code-level finding of the 2026-08-30 cold-start audit, across the CLI, API, dashboard, installer, and docs.

• FixedPrompt Injection, Skill Security, and Inference Security findings now render in the default text output
• Fixed--format json emits a single valid JSON document ({"summary", "findings"}); progress lines go to stderr
• Fixedsigil clone no longer flags its own shallow clone; PROV-006 no longer fires on plain directory scans
• FixedAPI: /forge/search handles fractional trust scores; GET /scans?scope=all no longer hangs; plan and credit rejections return 402/403
• Fixedinstall.sh surfaces glibc and checksum failures, falls back to the releases/latest redirect when rate-limited, honours SIGIL_VERSION
• ImprovedDocs: removed six documented-but-nonexistent commands; corrected verdict thresholds, exit codes, install methods, and phase counts

Registry statistics engine and trending analytics backend.

• AddedRegistry statistics engine with background processing across ecosystems, refreshed every 15 minutes
• AddedTrending analytics API and multi-criteria search sorting
• ImprovedDatabase-backed caching with 15-minute update intervals

False-positive remediation: clean repositories no longer score CRITICAL.

• FixedContext-aware pattern matching distinguishes JavaScript regex methods from shell execution
• FixedDocumentation and test files receive reduced severity; node_modules is skipped by default
• FixedUnicode boundary panics on multi-byte UTF-8 input
• AddedSafe-domains allowlist for known-legitimate API endpoints and a regression suite for false positives

Sigil comes to Claude Code. Scan repos, audit packages, and review quarantine without leaving your editor. Plus the full documentation site at sigilsec.ai/docs.

• AddedDocumentation site with guides for CLI, Claude Code Plugin, MCP, CI/CD, configuration, and troubleshooting
• AddedFour slash commands for Claude Code: /scan-repo, /scan-package, /scan-file, and /quarantine-review
• Added@security-auditor agent that explains scan findings and recommends fixes
• Added@quarantine-manager agent that guides you through approve/reject decisions
• AddedAuto-scan hooks — Sigil suggests a scan whenever you clone, install, or mention security
• AddedConfigurable auto-approve threshold so low-risk packages flow through without interruption

Install Sigil anywhere. Pre-built binaries for macOS, Linux, and Windows — plus Homebrew, npm, and Cargo.

• AddedPre-built binaries for macOS (Apple Silicon and Intel), Linux, and Windows — with SHA256 checksums
• AddedInstall your way: brew install, npm install -g, cargo install, or a one-line curl script
• ImprovedAutomated release pipeline publishes binaries to GitHub Releases on every tag
• FixedPackaging and installation fixes resolved across v1.0.1 through v1.0.4

Sigil is open source. Eight-phase scanning, quarantine-first workflow, threat intelligence, and MCP integration — all free under Apache 2.0.

• AddedOpen-source release of the Sigil CLI under the Apache 2.0 license
• AddedEight-phase scanner covering install hooks, code patterns, network exfiltration, credentials, obfuscation, provenance, prompt injection, and skill security
• AddedQuarantine-first workflow — nothing runs until you approve it
• AddedCloud threat intelligence with 55 detection signatures and a curated known-threats database
• AddedMCP server so AI agents can scan dependencies before installing them
• AddedEarly VS Code and JetBrains plugin support
• SecurityAll scans run locally — your source code never leaves your machine

Know who published a package and whether it has been seen before. Cloud threat intelligence, publisher reputation, and per-project ignore rules.

• AddedCloud threat intelligence enriches every scan with data from the Sigil community
• AddedPublisher reputation scores based on aggregated community scan data
• AddedThreat signatures stay fresh with automatic background sync (24-hour cache)
• Addedsigil diff command lets you compare scans over time and catch regressions
• Added.sigilignore file support to exclude directories and files from scans
• ImprovedFewer false positives on common environment variable patterns
• FixedShell alias setup now works correctly with Zsh and Oh My Zsh

See your scan history, manage your team, and browse threat intelligence — all from a web dashboard.

• AddedWeb dashboard with scan history, team management, and account settings
• AddedDetailed scan views with findings grouped by phase and severity
• AddedThreat intelligence browser with known threats, community reports, and detection signatures
• AddedTeam features: invite members, assign roles, and manage access
• AddedAlert channels for Slack, email, and webhook notifications
• AddedBilling and subscription management with plan selection

Cloud backend for Sigil Pro and Team plans. Scan storage, threat intelligence lookups, team management, and usage-based billing.

• AddedCloud API powering the Pro and Team plan features
• AddedScan history stored in the cloud — review past results anytime
• AddedThreat intelligence lookups by file hash
• AddedTeam management with invites, roles, and member removal
• AddedScan policies: set auto-approve thresholds, allowlists, and blocklists for your team
• AddedReal-time alerts via Slack, email, or webhook when high-risk scans complete

Run Sigil in your CI/CD pipeline. GitHub Actions, GitLab CI, and Docker support with SARIF output for code scanning.

• AddedGitHub Actions integration — add Sigil to any workflow with a single step
• AddedGitLab CI template for drop-in pipeline scanning
• AddedSARIF output for GitHub Code Scanning and other SARIF-compatible tools
• AddedDocker support with multi-stage builds and Compose for local development
• ImprovedContainer runs as non-root user for better security defaults

Sigil meets your editor. VS Code extension, JetBrains plugin, and an MCP server for AI agents.

• AddedVS Code extension to scan your workspace, individual files, or packages from the editor
• AddedJetBrains plugin for IntelliJ, PyCharm, and WebStorm
• AddedMCP server with 6 tools so AI agents can scan before they install

Scan URLs, archives, and MCP server configs. Plus baseline diffing to catch new risks between scans.

• Addedsigil fetch command to download and scan files from any URL
• AddedAutomatic archive extraction for .tar.gz, .zip, and .tar.bz2 files
• Addedsigil diff to compare scans against a baseline and surface new findings
• AddedDetection of MCP-specific risky patterns like auto_approve and allow_dangerous
• ImprovedBroader network exfiltration detection: Discord webhooks, Telegram bots, ngrok tunnels

Layer in your existing tools. Sigil now runs semgrep, bandit, trufflehog, and npm audit alongside its own scanner.

• AddedRun semgrep, bandit, trufflehog, safety, and npm audit as part of every Sigil scan
• AddedCloud threat intelligence — look up file hashes against a shared threat database
• Addedsigil login to authenticate and access Pro features

Get set up in seconds. Interactive installer, 9 shell aliases, and pre-commit hooks.

• AddedInteractive installer that configures shell aliases and hooks in one step
• Added9 shortcut aliases: gclone, safepip, safenpm, safefetch, audit, and more
• AddedPre-commit hook that scans every commit before it lands
• Added.sigilignore support for excluding files from scans
• SecurityPath traversal and input validation protections on all quarantine operations

Where it all started. Eight-phase scanner, quarantine workflow, and four-tier verdicts from LOW RISK to CRITICAL RISK.

• AddedEight-phase security scanner that catches install hooks, dangerous code patterns, network exfiltration, credential access, obfuscation, provenance, prompt injection, and skill security issues
• AddedQuarantine-first workflow — clone, pip install, and npm install all go to quarantine first
• AddedFour-tier verdict system: LOW RISK, MEDIUM RISK, HIGH RISK, CRITICAL RISK
• AddedDetailed scan reports with file paths and line numbers
• SecurityFully offline — no network calls, no telemetry, no data collection