Citation Corpora

Each ecosystem on this site maintains a shared citation corpus that all deep-dive articles in that series draw from. The corpus is verified once per release: every source has either a Semantic Scholar paperId recorded with a match score of at least 0.70, or it is a canonical non-S2-indexed reference (older paper, government standard, vendor specification) with a stable URL or DOI cross-checked manually.

The point of a shared corpus is twofold:

1. Coherence. When four articles in the same series cite Saltzer & Schroeder (1975), they cite the same DOI with byte-identical bibliographic formatting.
2. Auditability. A reader who wants to check the literature behind an article can read the source list once instead of repeatedly across articles, and a reviewer can verify every reference in one pass.

Wild Ecosystem corpus

Files: wild-citations.bib (BibTeX) · wild-citations.jsonl (per-source verification metadata with claim_anchors)

Covers: Wild Ecosystem hub · Part 1: Safety Architecture · Part 2: CLAUDE.md · Part 3: Observability Loop · Part 4: Claude Code as Tech Lead

Topics: capability-based access control, attribute-based access control, audit-log integrity, distributed-systems tracing, supervision trees, session types, two-phase commit, indirect prompt injection, language-model red teaming, long-context attention degradation, sycophancy, agent paradigms, software-engineering automation.

IRSB Ecosystem corpus

Files: irsb-citations.bib (BibTeX) · irsb-citations.jsonl (per-source verification metadata with claim_anchors)

Covers: IRSB Ecosystem hub · Part 1: On-Chain Enforcement · Part 2: Evidence Pipeline · Part 3: Watchtower Architecture · Part 4: AI-Agent Pivot · Guidewire MCP v0.1.0

Topics: Ethereum EIP-7702 delegation and ERC-7710 redemption, payment-channel watchtowers (Pisa, Cerberus, Outpost), smart-contract security and formal verification (Z3, Oyente, ZEUS, Hawk), Byzantine consensus and finality (Casper FFG, Tendermint, PBFT), MEV and frontrunning, off-chain computation patterns, cryptographic audit-log integrity, agent paradigms and prompt-injection threats, EU AI Act compliance, and enterprise-integration architecture (Parnas, Brooks).

Methodology

The verification pipeline reactivated for this work is documented at ~/.claude/skills/deep-research/references/semantic_scholar_api_protocol.md (v4.0, MCP-native). Each candidate source is resolved via mcp__semantic-scholar__paper_title_search or paper_details with a Levenshtein title-similarity threshold of 0.70 against the cited title; matches scoring above the threshold are recorded with the Semantic Scholar paperId and external identifiers (DOI, ArXiv, DBLP, MAG) in the .jsonl companion file. References that are not S2-indexed (e.g., older papers, vendor specifications, government standards) are flagged with verification_method: manual_* and a stable URL or DOI is recorded in their place.

This corpus does not claim to be a complete bibliography of the topics it covers. It is the working set of sources the deep-dive articles in this series cite. New sources are appended as needed; existing entries are not modified without bumping the entry’s verification timestamp.