Eda Knowledge Gate Maintainer

EDA Knowledge Gate Maintainer

When to use

Use this skill when the task needs:

1. repo experiment-hygiene gate execution,
2. enforced tool-reuse checks before adding scripts,
3. maintenance-log writeback during scoped execution,
4. diagnosis or repair of gate-hygiene failures.

Shared Capability Boundary

This skill owns:

• invocation of the repo knowledge gate,
• tool-query evidence for scoped execution work,
• maintenance-log discipline around substantive work,
• hard-failure policy for missing gate evidence.

It does not own:

• domain conclusions,
• workflow routing,
• artifact cleanup policy,
• KB retrieval beyond gate hygiene.

Expected Downstream Consumers

Typical consumers include:

• workflow-scoped-execution before and after substantive execution,
• execution skills that submit jobs, create scripts, or write experiment artifacts,
• eda-infra-maintainer when gate behavior itself needs repair.

Inputs

Provide or derive:

1. interaction scope,
2. task brief path,
3. keywords for tool query,
4. whether the call is routine gate use or gate-hygiene repair/debug.

Outputs

Return or update:

• gate command evidence,
• tool-query evidence,
• maintenance-log row in slurm_logs/00_meta/knowledge_tool_maintenance_log.md,
• explicit hard-failure reason if the gate contract is not satisfied.

Knowledge And Tool Interaction

1. Treat knowledge_gate.sh and tool_catalog.py query as canonical utilities, not optional suggestions.
2. Use the task brief and current maintenance log as the minimum local evidence set for gate execution.
3. Escalate to eda-infra-maintainer if repeated gate failures indicate a broken policy or missing infrastructure rather than a one-off task issue.

Hard Rules

1. Gate command failure is a hard stop.
2. Missing tool-query evidence for non-trivial work is a hard stop.
3. Missing maintenance-log update after file/script/job changes is a hard stop.
4. Do not silently continue if the task brief is missing for a scoped experiment interaction.

Operational References

1. Load references/invocation-scope.md when deciding whether the call is routine workflow-scoped-execution usage or direct gate-hygiene repair/debug.
2. Load references/gate-run-and-tool-query.md when running the gate command and tool-catalog query.
3. Load references/maintenance-log-writeback.md when writing the maintenance-log row and artifact summary.
4. Load references/hard-failure-policy.md when deciding whether the interaction must stop or escalate due to missing gate evidence.
5. Load references/checklist.md when you need the compact pre/post checklist view.