Repository navigation
Tags: savvides/idstack
Tags
v3.3.0.4 — doctor + readiness coverage, shared test helper Closes the May/June bot PR queue: 26 open PRs down to 0. No user-facing change. Adds execution coverage for bin/idstack-doctor (13 cases, previously none) and bin/idstack-status --readiness (22 assertions, previously none), and replaces nine drifted copies of the test check() helper with test/test-helper.sh. Test suites in CI 9 -> 11. smoke-test 371 -> 393 assertions, mutation suite 14 -> 18.
v3.3.0.3 — grep fallback --type filter never matched bin/idstack-learnings-search falls back to grep when python3 is absent. That fallback filtered on "type":"technical" while idstack-learnings-log writes "type": "technical" via json.dumps, so --type matched nothing idstack had ever written and returned zero results rather than filtering. Found by writing the first test that ever exercised the fallback, while deduplicating its two byte-identical copies. Also ships the last mergeable work from the May/June bot queue: first behavioral coverage for idstack-learnings-delete, -promote, -search flags, and three branches of idstack-timeline-log. integration-test 25 -> 48.
v3.3.0.2 — apostrophe search returned wrong results bin/idstack-learnings-search interpolated its arguments into Python source, so searching for a term containing an apostrophe (Bloom's, learner's) raised a SyntaxError that 2>/dev/null swallowed, fell through to a grep fallback with no --type filter, and returned a wrong-typed record as if correct. Also: bin/idstack-manifest-merge now rejects a non-object payload root instead of merging it and failing later inside idstack-status --readiness. From the stale Jules PR queue (#33, #48); 15 other PRs in that queue closed as superseded, duplicate, or incorrect.
v3.3.0.1 — skills no longer suggest commands that don't exist The no-bare-/skill guard added in v3.3.0.0 matched only backticked references, so three plain-prose examples in templates/preamble.md survived it. That preamble is spliced into all 22 skill files and those lines shape the model's welcome-back and next-step messages, so users were told to run commands like /assessment-design, which resolve in neither CLI. Fixed, guard widened to any command position, mutation case added.
v3.3.0.0 — course memory, pipeline orchestration, re-run detection Fixes a set of bugs that broke user-visible behavior, several since their features shipped: context recovery never ran on Python < 3.12, the pipeline orchestrator could not invoke child skills, re-run detection was dead in five skills, and standalone runs of three skills silently never persisted results. Adds the CI that would have caught them, the first tests for ./setup, and a mutation suite that proves each guard fails when its bug returns. See CHANGELOG.md for the full entry.
PreviousNext