Where Accountability Lives: source collection of vendor documentation and provider terms for four agentic coding tools

Archived source collection underlying the paper Where Accountability Lives: Mapping Human Responsibility to Workflow Artifacts in Agentic Software Development. The collection comprises 121 manifest items retrieved on 16 August 2026, of which 118 returned HTTP 200 and were archived and 3 returned HTTP 404 and are recorded as unreachable with the URLs attempted. It covers the product documentation of four agentic coding tools (GitHub Copilot cloud agent, Devin, Cursor, Claude Code), the platform controls that operate on agent output, and 18 governing policy documents across 8 providers. Every archived item carries a provenance header recording the requested URL, the served URL where these differed, the HTTP status, the page title parsed from the served markup, any date string present in that markup, the SHA-256 of the body as served, the byte length, and the retrieval timestamp and method. Response bodies are additionally stored byte-exact without a header, so any extraction can be re-checked against the untouched response. Twenty-three items were served from a URL other than the one requested; both URLs are recorded. Six extraction records accompany the archive, one per document group, containing the passages returned by the extraction rules together with the rules themselves. A separate absence record lists, per policy document, every search term and its hit count, so that a claim of absence can be judged against the search that produced it. No language model was used in the retrieval or extraction path. Both are performed by the deposited scripts, which issue direct HTTP requests, write response bodies byte-exact, and apply lexical rules. The collection is reproducible from the manifest without any model. The deposit also includes the survival test used to measure the reliability of absence claims, the consistency checker that verifies the manuscript's counts against the retrieval index, the reference verifier that checks citations against the arXiv API, and a blind re-coding harness for independent classification. Two limitations are stated in the README and repeated here. The extraction rules under-detect: they are lexical, and at least one page containing merge-queue language was not matched by them. Absence is relative to the pages searched: an earlier version of this work covered seven pages of one vendor's documentation and reported an absence that a wider sweep overturned. Every absence in the paper was confirmed by reading the archived body rather than by the absence of a pattern match. Archived documents remain the property of their publishers and are reproduced for verification of the analysis. The manifest, scripts, extraction records, and retrieval index are released under CC BY 4.0.