dwbimstr/context-foundry is a Rust context engineering toolkit that provides coding agents with citation-based retrieval, graph context, explicit memory, and bounded token delivery. The 2026-10-27 (UTC) commit 69c3bb0 merged a security fix that bumps this repository's Markdown parser library pulldown-cmark from 0.13.4 to 0.13.5, along with the accompanying unit move. The trigger was a bug where 0.13.4's default parser terminated with 101 on a single 13-byte Markdown list-and-indentation input, and the per-source panic handling at the indexing stage only leaves that source in a parse-failure state, failing to block the later outline/context re-parsing.
Because a change in Markdown parsing results changes the very source units this tool cuts apart, the commit raises the search schema to 8 and SOURCE_UNIT_SCHEMA to 6. The repository schema 12 and generation v6 stay as they are, but existing stores must explicitly run foundry --store STORE repair-index to produce search schema 8 and then, in the following semantic prepare, re-align to source-unit schema 6. The change amounted to 1,262 added and 74 deleted lines across 19 files, and the commit message states1,314 passed, 0 failed, and 34 ignored against the local gate.
Some things were held back as well. tree-sitter-scala 0.26.3, rmcp 3.7.0, and uuid 1.29.0 were not bumped in this commit, and the Scala grammar change was also left on hold because it can change the units. The documentation records this work as a security-only measure approved as an exception in the G2 promotion measurement, claims no quality or performance improvements, and states that the complexity fix requires upstream intervention, so no time was spent on it here.
There are limits to the verification. The code was written by GLM and reviewed by GPT-6.1 Sol and GPT-6 Astra, but it explicitly states that the two models belong to the same supplier and therefore are not counted as two independent verifications. Those working with an already built store have to accept the two steps of repair-index and semantic prepare, while those starting fresh do not have to go through the migration steps.