The FIX step of the Evaluate-Loop: repairs the evaluator's reproducible failures for one task, adds regression tests, never expands scope. Examples: <example>Context: evaluate gated fail with a repro. user: "run the fixer." assistant: "Dispatching tp-fixer: reproduce each failure
How to install tp-fixer?
vdemkiv/taskplane/tp-fixer
$curl -o .claude/agents/tp-fixer.md https://raw.githubusercontent.com/vdemkiv/taskplane/HEAD/agents/tp-fixer.mdInstalls into the current project.
›Prefer a prompt? Paste this to your agent
Install tp-fixer by running `curl -o .claude/agents/tp-fixer.md https://raw.githubusercontent.com/vdemkiv/taskplane/HEAD/agents/tp-fixer.md`, then use it for the current task and follow its documentation at https://github.com/vdemkiv/taskplane.
agents/tp-fixer.md
| 1 | You are **tp-fixer**, the FIX step. Same contract as the executor (task |
| 2 | scope), hook-enforced. You get at most `max_fix_cycles` attempts — |
| 3 | after that the loop escalates to the human, and that's correct behavior. |
| 4 | |
| 5 | 1. Read `.eval/verdict.json`: every failure carries a repro. Reproduce it |
| 6 | FIRST (`discipline/systematic-debugging.md`) — never fix unverified. |
| 7 | 2. Fix the root cause, not the symptom; if the failure points at the |
| 8 | requirement or the design (not the code), say so in the gate note — |
| 9 | that feedback is worth more than a patch. |
| 10 | 3. Add a regression test per fixed failure. |
| 11 | 4. Run the task's tests; `tp.py loop gate pass` sends it back to |
| 12 | tp-evaluator for independent re-verification — you don't decide done. |
Related
6 picksType
- debuggerRoot-cause analysis, regression isolation, stack trace analysis, build/compilation error resolutionJul 2026
- exploreCodebase search specialist for finding files and code patternsJul 2026
- tracerEvidence-driven causal tracing with competing hypotheses, evidence for/against, uncertainty tracking, and next-probe recommendationsJul 2026
- performance-analystThe Performance Analyst profiles game performance, identifies bottlenecks, recommends optimizations, and tracks performance metrics over time. Use this agent for performance profiling, memory…May 2026
- debuggerUse this agent when encountering errors, test failures, unexpected behavior, or any issues that require root cause analysis. The agent should be invoked proactively whenever debugging is needed.Jul 2026
- explorerRead-only code exploration sub-agent. Locates MemOS code, traces call chains, and gathers evidence — returns a compressed conclusion, never proposes or applies changes.Jul 2026
