EuLogikon for AI agents

Name: Eulogikon Ancient Greek Text Corpus
Creator: Eulogikon
License: https://creativecommons.org/publicdomain/mark/1.0/

The full corpus — 1,341 authors and 3,998 works of ancient Greek — is published as a machine-readable agent surface. The texts are public domain; Eulogikon claims and waives nothing. Stable identifiers, byte-verified retrieval, no authentication.

What this is and isn’t. This surface targets agent runtimes: npx skills add, agentic IDEs and CLIs (Claude Code, Cursor, Codex, Continue, Cline, OpenCode and others), and custom pipelines. Chat assistants like Claude.ai and ChatGPT do not yet auto-discover well-known agent skills from arbitrary domains — that capability is expected to land as web-discovery matures. If you pasted this URL into a chat assistant and it didn’t fetch anything, that’s why.

One-line install

For runtimes that implement the Agent Skills Discovery RFC v0.2.0 (Claude Code, Cursor, Codex, Continue, and others — see the skills package’s keyword list):

npx skills add https://eulogikon.org

The client fetches the discovery index, verifies the SKILL.md digest, and installs the skill into whichever runtime it detects. From then on, the agent has structured access to authors, works, and individual canonical JSON files.

Try it in three lines

Plato’s Apology of Socrates is eul_wid ffk-aa. The opening line lives at Stephanus reference Apol 17 a. Verifying the bytes:

# 1. look up the manifest entry
$ curl -s https://eulogikon.org/data/works-manifest.json \
    | jq '.works[] | select(.eul_wid == "ffk-aa")'
{
  "eul_wid": "ffk-aa",
  "url": "https://eulogikon.org/data/works/grc/plato-apology-ffk-aa.json",
  "sha256": "2dae06fd4b14890b5638c85c487d945cb18b27608177d80f1dec6ecd6d17267d",
  "bytes": 115257
}

# 2. fetch the work and verify the bytes match
$ curl -sL https://eulogikon.org/data/works/grc/plato-apology-ffk-aa.json \
    | shasum -a 256
2dae06fd4b14890b5638c85c487d945cb18b27608177d80f1dec6ecd6d17267d  -

# 3. read the opening line, addressed by Stephanus reference
$ curl -sL https://eulogikon.org/data/works/grc/plato-apology-ffk-aa.json \
    | jq -r '.units[0] | "\(.ref): \(.text)"' | head -c 200
Apol 17 a: Οὕτι μὲν ὑμεῖσ, ὶ ἄνδρεσ Ἀθηναῖοι, πεπόνθατε ὑπὸ τῶν ἐμῶν κατηγόρων…

The page for human reading is https://eulogikon.org/grc/plato/apology. The eul_wid, the sha256, the Stephanus reference, and the URL all stay stable across releases.

Markdown variants

Every work and author page has a .md companion at the same URL. Two ways to get it:

# Direct fetch — the .md companion at the same URL
curl https://eulogikon.org/grc/plato/apology.md

# Content negotiation against the clean URL
curl -H "Accept: text/markdown" https://eulogikon.org/grc/plato/apology
# → 308 redirect to /grc/plato/apology.md

The .md file carries YAML frontmatter (eul_wid, canonical URLs, sha256, classification, license) and a body with one ### H3 per unit ref — a stable citation anchor map. HTML stays the default for browsers and clients that don’t signal a markdown preference. Markdown is typically ~40% the size of the HTML for the same content.

Direct fetch

/.well-known/agent-skills/index.json Discovery index (RFC v0.2.0). One skill, pointing at the SKILL.md.

/.well-known/agent-skills/eulogikon-corpus/SKILL.md Full instructions: identifiers, workflows, citation format, what’s in each data file.

/data/works-manifest.json Manifest of every canonical work JSON with sha256 + byte length. Entry point for bulk crawl or verified single-work fetch.

/data/authors.json 1,341 authors keyed by eul_aid, with period, dialect, affiliation, era, floruit dates.

/data/works.json 3,998 works keyed by eul_wid. Most filter queries can be answered single-pass from this one file.

/llms.txt Apex orientation for AI-targeted crawlers (llmstxt.org convention).

Conventions

Identifiers are stable. eul_aid identifies authors (lowercase slug, e.g. plato). eul_wid identifies works in the form {eul_aid}-{suffix} (e.g. plato-aa). Both are corpus-wide and permanent.

Page URLs follow the grammar https://eulogikon.org/grc/{author_slug}/{work_slug}. Slugs are stable; author_slug and work_slug are first-class fields in works.json so an agent never needs to guess.

Within-work addressing via TLA fragments — sid for a point, rid for an inclusive range — is currently deployed for the Hippocrates corpus only. Corpus-wide rollout is in progress. Where deployed, RID fragments can be appended to work URLs (#aaa-aaa for a point, #aaa-aag for a range) and the full grammar is documented in the SKILL.md.

License and use

Eulogikon publishes public-domain Greek texts in cleaned, structured, readable form and makes no copyright claim over that preparation. Every surface carries Public Domain Mark 1.0. No attribution required. No restrictions on use, including training. Citations via eul_wid plus the sha256 from the manifest give verifiable provenance if you want it.

EuLogikon does not attribute Greek texts to any institution. The texts are the texts.

Contact

Corrections, questions, or suggestions are welcome at /contact. If a digest doesn’t match or a manifest entry looks wrong, please flag it — integrity of the byte-level claims is load-bearing for everything this surface promises.