Strip trailing _query_source column before header validation

[Morabbin]

Summary

session_store_sql supplements results from the local session store with a trailing _query_source column. materialize-session-query.py validates result headers against an exact expected column list, so any result carrying that column was rejected with:

unexpected discovery result column _query_source

This blocked Repository Skill Forge discovery whenever the underlying session_store_sql result came from (or was supplemented by) the local store.

Context: github/agents#1815.

Scope note: this fixes a local/store result-compatibility bug in the materializer's header validation. It does not address the production seven-day scale issue tracked separately; those are unrelated problems that happened to surface through the same script.

Change

In table_rows(), drop a trailing _query_source column from the header before the per-kind header equality check runs, and strip the corresponding cell from each row with a single right-hand split (row.rsplit(" | ", 1)[0]):

if header and header[-1] == "_query_source":
    header = header[:-1]
    rows = [row.rsplit(" | ", 1)[0] for row in rows]

This only fires when _query_source is genuinely the last header column, so:

• Results without the column are unaffected.
• Unknown extra columns are still rejected by the unchanged header equality check.
• A _query_source column that isn't trailing is still rejected (it doesn't match the expected header once the intended trailing check finds a different last column).
• The right-hand split removes only the final Markdown cell, so embedded " | " sequences in earlier values (e.g. arguments_json) are preserved.

Merge conflict note

origin/morabbin-forge-deterministic-worker refactors result_content() into probe_result() / materialize_content() in this same file, but it does not touch table_rows() or parse_rows(). This PR is based on main and targets only table_rows(), so a conflict is unlikely; if one does occur it should be limited to nearby context lines, not overlapping logic.

Tests

Added to tests/test_materialize_session_query.py, exercising table_rows() and parse_rows() directly:

• No _query_source column (baseline, unchanged behavior).
• Trailing _query_source column stripped and parsed correctly.
• Unknown trailing column (not _query_source) still rejected.
• _query_source present but not trailing still rejected.
• Embedded " | " in arguments_json preserved through stripping (tool-calls kind).
• Zero rows with a trailing _query_source header.
• Malformed row (missing the _query_source cell) raises downstream in parse_rows instead of silently misparsing.

Validation

$ python3 -m unittest tests.test_materialize_session_query -v
Ran 26 tests in 0.100s
OK

$ python3 -m unittest discover -s tests -p "test_*.py" -v
Ran 105 tests in 2.012s
OK

$ python3 -m py_compile scripts/materialize-session-query.py tests/test_materialize_session_query.py
$ python3 -m compileall -q plugins/repo-dreamer/skills/repository-skill-forge
(no errors)

No repo-wide lint/format tooling is configured for this repository, so none was run.

[Copilot]

Pull request overview

Updates session query materialization to accept local-store results containing a trailing _query_source column.

Changes:

• Removes trailing _query_source headers and corresponding row cells.
• Adds coverage for valid, malformed, embedded-pipe, and unexpected-column cases.

Show a summary per file

File	Description
scripts/materialize-session-query.py	Normalizes supplemented query results before schema validation.
tests/test_materialize_session_query.py	Tests normalization and rejection edge cases.

Review details

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

• Files reviewed: 2/2 changed files
• Comments generated: 0
• Review effort level: Balanced