1Panel

The evidence provides a vulnerability-reporting address, a supported-version statement, backup and restore claims, and one workflow restricted to pull-requests: write. Repository naming, community links, and a maintenance contact provide basic attribution. Deductions apply because no least-privilege model for Agent or server operations, confirmation gates for hazardous actions, data-flow inventory, credential lifecycle, or secrets-handling policy is shown. Installation relies on an uninspected remote script, while workflows use third-party Actions and custom tokens without evidence of dependency scanning, commit-digest pinning, or supply-chain mitigation. External effects are described but their authorization boundaries and safeguards are not. Backup and restoration merit a higher score, although scope, validation, and failure behavior are undocumented. Unknown publisher identity is not itself penalized, but the supplied attribution evidence is not thorough enough for full marks.

The README is broadly internally consistent about server management, AI management, product tiers, and support routes, while SECURITY.md clearly limits support to v1.x. Deductions apply because the main installation path depends on a live remote script, external resources, and third-party Actions without mirrors, checksums, a compatibility matrix, or availability fallbacks. No error-message examples, diagnostic guidance, or failure-recovery instructions appear, so failure_messages receives zero. Missing runtime testing was not used to reduce these static criteria.

The material identifies Linux self-hosters and covers website, container, database, file, backup, and AI-management scenarios. The OSS/Pro/Ent table gives reasonably clear commercial capability boundaries. Deductions apply because no Agent trigger conditions, routing logic, false-trigger controls, or tool-selection rules are supplied, leaving trigger_precision unsupported. Environment guidance stops at a generic Linux server and access URL, without distributions, architectures, resource requirements, ports, container-runtime prerequisites, or network constraints.

The README has clear sections for overview, features, quick start, editions, support, security, and licensing, plus many language entry points; 1Panel naming is stable in the supplied material. GPL-3.0 metadata, the README statement, and the full LICENSE agree, justifying full license marks. Deductions apply because installation is essentially a curl-to-shell command without verification, prerequisites, or uninstall guidance; there is a screenshot but no Agent example or FAQ; stated limitations mostly concern paid tiers rather than technical constraints. A releases link and the v1.x support table establish an update path, but no actual changelog content is provided. Issues, Discord, and the security email show maintenance channels, although publisher registry identity is unverified and ownership responsibilities are not detailed.

The repository describes a unified web GUI, one-click deployment, an app store, backup and restoration, and AI management, while the edition table communicates free and paid choices. This supports some prospective usability and cost comparison. Deductions apply because no concrete Agent inputs, outputs, workflows, or resulting artifacts are shown, so direct output usability cannot be established statically. The competitor table lacks methodology or supporting evidence, making marginal-value claims largely promotional. OSS availability and tier differences are clear, but hardware, operations, model usage, storage, and migration costs are not disclosed.

The license, supported version, security contact, and installation command are directly traceable to supplied files, and README, LICENSE, and SECURITY.md provide limited corroboration on those matters. Deductions apply because major claims such as 2.5M+ users, 165+ applications, enterprise-grade security, reduced vulnerability exposure, and competitor advantages lack accompanying static substantiation. AI Agent functionality is represented only by overview and tier statements. Facts, inferences, and promotional language are not systematically separated, while cross-file corroboration is largely confined to licensing and maintenance routes.