FULL-REPOSITORY SOLANA AUDITS
SPECTRE combines programmatic analysis with grounded model adjudication to trace findings across supported files and programs. A full audit turns that evidence into remediation guidance, a patch and regression proof where suitable, and a complete audit pack.
Find the issue · Trace the path · Prove the fix
Product preview of a SPECTRE finding with connected source context, regression evidence, a verified patch state, and the full audit pack.
From scan to solution
Programmatic analysis creates repeatable coverage. Repository context and grounded adjudication explain what matters. Full-audit delivery gives engineering the evidence and the next move.
Solana-aware programmatic analysis scans supported source, accounts, instructions, authorities, dependencies, and cryptographic use. Coverage begins with code analysis—not a prompt.
SPECTRE follows evidence across files and program boundaries so a finding arrives with the surrounding code that gives it meaning.
A grounded model evaluates the cited source and relevant relationships, recording a verdict, rationale, confidence, and provenance. Uncertain candidates stay visible.
A full audit provides concrete remediation guidance and can include a patch plus regression proof that fails before the fix and passes after it.
The model adjudicates. The evidence remains inspectable.
Separating candidate discovery from judgment makes the scan repeatable while preserving the contextual reasoning needed for protocol risk. Uncertain results fail open for review.
One full audit
A SPECTRE full audit gives engineering enough context to act and a portable audit pack that security, governance, and procurement can carry forward.
Finding delivery
The rule, exact source location, rationale, and relevant paths stay attached to the finding instead of collapsing into an alert ID.
Every delivered finding explains the next engineering action. Suitable full-audit findings can also include a ready-to-apply patch.
Where executable verification is appropriate, the reproducer fails before the fix and passes after the remediation.
Complete audit pack
Detected, adjudicated, reproduced, and patched are separate evidence states. SPECTRE only presents the stronger state when the supporting evidence exists. Patch and test artifacts accompany suitable findings; the audit pack remains the stable record of the analysis.
Why SPECTRE
Conventional SAST provides repeatable detection. LLM-first review provides flexible reasoning. SPECTRE connects programmatic analysis to grounded adjudication and a full-audit workflow built around engineering action.
Swipe horizontally to compare approaches →
| Approach | Conventional SAST Broad, repeatable coverage of known patterns | LLM-first review Exploratory review and novel hypotheses | SPECTRE Repository audit with a path to closure |
|---|---|---|---|
| Candidate discovery | Rules, queries, control flow, and data flow | Model reasoning over supplied or gathered context | Programmatic Solana and cryptography analysis |
| Repository context | Advanced products can reason across files and functions | Can gather broad repository context, but remains probabilistic | Relevant relationships across supported files and programs |
| Model role | Usually absent or used for enrichment | Primary detector and reviewer | Grounded adjudication after candidate discovery |
| Typical output | Alerts, traces, and remediation guidance | Review comments, explanations, and suggested changes | Finding, evidence state, remediation path, and full audit pack |
| Fix proof | Before/after regression proof is not a standard deliverable | Suggested changes still require independent verification | Patch and before/after regression proof where suitable |
Category capabilities vary by vendor and deployment. SPECTRE is designed to complement—not replace—manual review for novel protocol behavior, runtime conditions, and economic invariants.
Full-repository scope
SPECTRE looks beyond isolated findings to the supported source, state, authorities, dependencies, cryptography, and program relationships that determine real impact.
Review supported account constraints, signers, PDA derivation, CPI safety, state invariants, and instruction behavior with Solana-specific meaning.
Follow relevant definitions, handlers, authorities, calls, and program boundaries instead of judging an isolated call site alone.
Resolve software inventory and dependency relationships alongside the source paths and program assumptions that rely on them.
Inventory primitives, libraries, key-material use, and authorities, with CBOM output and migration-impact context attached to the audit.
Cryptography is one first-class audit surface, not a separate product. SPECTRE can emit a CycloneDX CBOM and standards-linked migration-impact context, but protocol-layer replacement choices remain with the relevant standards, runtimes, and engineering teams.
Supported source
A Solana repository is rarely one language. SPECTRE extracts the on-chain program and its off-chain clients as one graph, then extends that same analysis to whatever else lives in the repository.
Tree-sitter extraction over programs, instruction handlers, Accounts structs, constraints, and PDA derivation, checked against the Solana rule pack for governance, authority, and CPI safety.
Client call sites, SDKs, and test suites are extracted and bound back to the on-chain instructions they invoke, so a finding carries both sides of the program boundary.
Also detected in mixed-language repositories
Language detection is automatic. When a repository mixes languages, SPECTRE processes each of them and merges the results into a single unified graph rather than scanning files in isolation.
Your repository, with a way forward
SPECTRE for Solana is rolling out to select teams. Request a full-repository audit or early access to the unified dashboard.
