agent-kickstartClaude Code onboarding
A guided, project-local first experience for Claude Code beginners.Evidence boundary: Candidate entry point; fresh-machine, cross-platform, and privacy review remain open.
Open-source AI agent reliability tools14-project public catalog
The Hermes Labs open-source tools catalog.
The Hermes Labs public catalog is a workflow, not a wall of repositories: begin with a project-local setup, inspect language and input data before runtime, observe and guard live behavior, evaluate changes, and preserve the context behind decisions. Each tool below names the boundary of what its present evidence supports.
Follow the path: start → static checks → runtime → evaluate → memory
See also: research and evidence · upstream contributions
Try a real LintLang scan in your browser →
Evaluating these controls for production? Discuss adoption and evidence boundaries →
§ 01
Curated public catalog
Tool · role
Purpose · evidence boundary
Try · install · source
StartOne guided entry point
Static checks & drift gatesStatic analysis, data preflight, and drift gates
lintlangRelational agent-language linter
Deterministic static analysis for agent instructions, including H1.6 pairwise differentia: comparisons of tool descriptions within one parsed input that report when one or both lack distinguishing wording. Zero LLM calls.Evidence boundary: H1.6 uses a finite term-and-synonym model and does not establish runtime tool-selection improvement; a clean scan does not prove runtime safety.Public evidence: 3,800+ PyPI registry downloads through 13 August 2026 · Tool Differentia DOI · Tool Differentia PDF · recurring CI in Character.AI's public Larch repository · Larch's linting reference linking lintlang as the upstream · independent Gentoo packaging in the unofficial Haven overlay · attributed product influence on Agent Lint
csv-quality-gateCSV preflight quality gate
Checks a CSV for missing required columns, empty files, empty critical cells, duplicate rows, and obvious junk values before a pipeline ingests it, returning pass, warn, or fail with matching exit codes. Stdlib-only; runs as a CLI, a pre-commit hook, or a GitHub Action.Evidence boundary: Shape and obvious-noise heuristics on the columns a profile names; not semantic validation of the values and not a data platform with lineage, profiling, or schema inference.
hermeneuticCorrection-aware drift gate
Mines corrections from chat logs (prior response, user correction, repair) and runs a preflight gate on the next response.Evidence boundary: A drift check, not proof that the proposed response is correct or complete.
rule-auditStatic system-prompt analyzer
Parses an AI system prompt into normative rules and reports logical contradictions, coverage gaps, priority ambiguities, meta-rule paradoxes, and absolute-rule edge cases, with CLI exit codes for CI and a pre-commit hook. Pure Python, no LLM calls, no network.Evidence boundary: Lexical parser, not a language model: implied or narrative-embedded rules can be missed, and a CRITICAL label is a lexical score, not a verified exploit.Public evidence: pre-commit hook merged by the mcs-agent-analyser maintainer · CI enforcement merged by the mcs-agent-analyser maintainer
Observe and guard at runtimeExperimental probes, policy monitoring, and a delete guard
little-canaryExperimental prompt-injection preflight
Probes untrusted input with a sacrificial canary model before the primary model acts.Evidence boundary: Experimental security tool; efficacy and false-positive behavior depend on the models and evaluation distribution.Public evidence: Behavioral Canarying DOI · Behavioral Canarying PDF
agent-gorgonRuntime policy guard
Scores observed file, network, and subprocess behavior against policy and records forensic evidence.Evidence boundary: Reactive user-space monitoring; it is not a sandbox or a prevention guarantee.
agent-trash-guardRecoverable-delete guard for coding agents
A pre-tool hook for Claude Code, Codex, and Gemini CLI that blocks recognized permanent-delete shell commands and redirects the agent to a recoverable trash CLI with list and restore commands. Installs from the repository checkout as a local plugin; no live indexed marketplace route yet.Evidence boundary: Blocks recognized permanent-delete shell commands such as rm, unlink, shred, rmdir, find -delete, and git clean -f; it does not cover overwrites, truncations, or cloud and API deletion, and it is not a sandbox.
agent-signagePre-tool git-state signs for coding agents
A PreToolUse hook for Claude Code, usable from any harness that can shell out before a file operation, that checks measurable git facts about the file an agent is about to touch (stale checkout, symlink outside the repo, another worktree mid-edit) and injects one short line into the agent's context only when something matters. Requires git on PATH; no package dependencies.Evidence boundary: Passive by contract: it never blocks, never exits non-zero, and goes quiet on every error, so it is not a gate; measurements contact no network, and the only network call is a detached, never-awaited background git fetch that can be disabled.
Evaluate and reviewEvidence-first scoring
hermes-rubricEvidence-first review protocol
Builds structured rubrics, binds scores to cited evidence, and records reproducibility receipts.Evidence boundary: Advisory, not a binary release gate; the published agreement result still needs independent reproduction.
Memory & recoveryLocal memory, session recovery, and a compact reference track
langstateInspectable context compression
Compresses older conversation turns into visible scaffold state and records whether explicitly named literal facts survived.Evidence boundary: Lossy compression prototype; its receipt checks literal text and does not establish semantic equivalence.
Fidelis MemoryLocal-first agent memory
Returns your original stored passages verbatim, with no LLM call in the default retrieval path.Evidence boundary: Published on PyPI as fidelis-memory; the repository, import name, and CLI remain fidelis. Independent install-to-recall proof remains open.Public evidence: source-referenced feature comparison of fidelis 0.0.9 by the unaffiliated ai-memory-comparison project (28 May 2026)
hermes-blindSession recovery scaffold
Extracts turn-one goals from Claude Code or Codex session logs into a compact anchor for long-task reorientation.Evidence boundary: Deterministic mission representation, not automatic drift detection or proof of improved model behavior.
zer0dexMemory reference implementation
A compact dual-layer memory pattern combining a human-readable index with vector retrieval.Evidence boundary: Teaching and reference track; benchmark implications are not independently reproduced.
§ 02
How to read this catalog
01
Inspect the evidence
Repository-level truth
A public repository is an inspectable artifact, not a claim of production fitness. Read its README, limitations, tests, and release history against your actual environment before adoption.
02
Separate kinds of proof
Local · live · independent
Local tests, a public package, a benchmark, and independent evaluation answer different questions. The boundary on each tool names material work that remains before a stronger claim is fair.
03
Use the research as context
Concepts · methods · limits
The research index provides the conceptual and empirical work behind Hermes Labs. It informs the engineering program; it does not certify every tool or deployment.
To reproduce a boundary, report a false positive, or improve an adapter, open an issue in the relevant project repository.
The wider organization remains available for historical and secondary source inspection. Those repositories are not part of this curated catalog. Browse the full GitHub organization