run(weekly/prompt-fixer-agent): automate prompt safety fixes

- Selected `prompt-fixer-agent` from weekly roster.
- Parsed `artifacts/prompt-safety-audit.md` identifying 24 agents needing fixes.
- Applied `FAILURE MODES` and no-op exit criteria to all 24 agent prompts.
- Generated required run artifacts (Context, Todo, Decisions, Test Log).
- Verified artifacts and linting.
- Published completion lock.

Co-authored-by: PR0M3TH3AN <[email protected]>

src/decisions/DECISIONS_2026-02-15T05-06-49Z.md

@@ -0,0 +1,22 @@
+---
+agent: prompt-fixer-agent
+cadence: weekly
+run-start: 2026-02-15T05:03:00Z
+prompt: src/prompts/weekly/prompt-fixer-agent.md
+---
+# Decisions - Prompt Safety Fixes
+
+## Decision: Automation Strategy
+- Used a Python script (`fix_prompts.py`) to parse the audit report and apply fixes.
+- This ensures consistency and reduces manual error for 24 files.
+
+## Rationale
+- **Failure Modes**: Appended to the end of the file (or before `OUTPUTS` if present).
+- **No-op Logic**: Inserted into `MISSION`, `WORKFLOW`, or `GOALS` sections. The script looks for these headers and inserts the line `- If no work is required, exit without making changes.` at the end of the section (before the next header).
+
+## Manual Intervention
+- `protocol-research-agent.md` required manual intervention because the script failed to identify a suitable insertion point with the initial logic (though updated logic might have worked, manual fix was safer and faster).
+
+## Verification
+- Verified that `const-refactor-agent.md` contains the new sections.
+- Verified that `protocol-research-agent.md` contains the new no-op logic.