trustsight 0.2.0

CLI-based AUR package update vetting tool

TrustSight

Audits AUR PKGBUILDs before you update: catches careless malice and structural risk, and tells you what it can't verify.

---

The 30-second example

trustsight review

                        TrustSight Review
┏━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ Package         ┃ Risk Score   ┃ Verdict                                   ┃
┃                 ┃              ┃                                           ┃
┡━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ chez-scheme-bin │   0/100      │ Version bump. No structural changes.      │
├─────────────────┼──────────────┼───────────────────────────────────────────┤
│ sketchy-pkg     │  35/100      │ R004 HIGH  Checksum disabled (SKIP).      │
│                 │              │ C003 INFO  Source URL changed without     │
│                 │              │            version bump.                  │
│                 │              │ SOURCE_BUCKET MEDIUM  New domain:         │
│                 │              │   sketchy-cdn.invalid (unknown).          │
│                 │              │ NOVELTY HIGH  Source URL first seen       │
│                 │              │   globally.                               │
│                 │              │ PINNING INFO  Source pinning: unpinned.   │
│                 │              │ Verdict: Checksum disabled; sources       │
│                 │              │   replaced with content from an unknown,  │
│                 │              │   never-before-seen domain.               │
├─────────────────┼──────────────┼───────────────────────────────────────────┤
│ obsidian-beta   │  15/100      │ INCONCLUSIVE. Only 2 prior observations;  │
│                 │              │ no high-severity signals from a cold DB.  │
└─────────────────┴──────────────┴───────────────────────────────────────────┘

The tiered evidence display is the differentiator: every signal (rule, bucket, novelty, pinning, verification) is shown with its contribution and severity. You see why the score is what it is.

---

What it catches / what it can't

Detected by TrustSight	Outside TrustSight's scope
Careless malice: curl | bash, base64 | sh, wget pipe sh (R001 recall ~100%). Obfuscated casing, embedded URLs in function bodies.	Signed upstream payload: the PKGBUILD is not the binary. A benign build file can fetch a tampered release tarball.
Structural risk: checksums disabled (R004), checksums emptied (R005), URL typosquatting (githab.com), source URLs swapped without a version bump (C003).	Deliberately-unremarkable PKGBUILDs: no added commands, no new URLs, no checksum changes; no signal. The update is invisible to diff analysis.
Anomaly-vs-history: first-seen URLs (global or per-package), first-seen maintainer, low-observation-count gating with INCONCLUSIVE verdict.	Build-dependency attacks: a malicious makedepends or depends is outside PKGBUILD scope. TrustSight audits the recipe, not the supply chain's second-order dependencies.
Reviewer manipulation: Unicode bidi overrides (R013, 88% recall) that make displayed text differ from executed text. Prompt injection in comments/descriptions (R012, 17% recall; kept as tripwire; primary defense is verdict-integrity assertions).	Unpinned sources result in INCONCLUSIVE. A source=($pkgname-$pkgver.tar.gz) with no checksum, tag, or commit pin is reported as structurally weak, not silently accepted.

---

Install

pip install trustsight

AUR: trustsight (dogfood: TrustSight audits its own updates).

From source:

git clone https://github.com/emiliano-go/trustsight
cd trustsight
pip install -e .

Requires Python 3.12+.

---

Commands

Command	What it does
trustsight review	Scan all outdated AUR packages and produce a scored table with tiered evidence.
trustsight inspect <package>	Deep-dive on a single package: full score breakdown, source URLs, resolved commands, novelty context.
trustsight history <package>	Show past analysis results for a package, with optional --score-breakdown.

---

How scoring works

Scoring is deterministic: same input always produces the same score. A core of 13 detection rules (R001 to R013) and 3 code-structure rules (C001 to C003) produces signals across four evidence tiers: A (structural, rules on PKGBUILD lines), B (priors/context, URL classification and forge trust), C (history/novelty, first-seen URLs and maintainers gated by observation count), and D (verification, checksums, PGP keys, GPG verify, which subtract from the score). The LLM is entirely optional and never calculates; it translates the deterministic breakdown into English, and verdict-integrity assertions gate its output. See scoring-philosophy.md.

---

Security model

TrustSight is evidence-producing, not proof-of-safety. It audits and does not install. The tool never runs the PKGBUILD, never executes extracted commands, and never modifies your system. Every finding is traceable to a specific diff line, URL, or novelty record. The output is a structured risk assessment to inform your decision, not a gate. See trust-model.md.

---

License

MIT. Deliberately permissive to encourage adoption, auditing, and fork-investigation by the Arch Linux and security communities.

---

Documentation hub

Section	Description
Getting Started	One-tutorial path from install to first review
Full documentation	Docs landing page
Contributing	How to report bugs, contribute code, improve docs
Security	Vulnerability disclosure policy
License	MIT full text