You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Jul 4, 2026. It is now read-only.
Introduce manifest-driven workflow gating, new solution audit capabilities, and stress profile execution fixes while aligning docs/tests with the expanded 20-tool surface.
Made-with: Cursor
Copy file name to clipboardExpand all lines: agents/autocode-workflow.md
+22-4Lines changed: 22 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -3,6 +3,11 @@ name: autocode-workflow
3
3
description: Coordinates AutoCode problem creation and enforces the full validator-generator-checker workflow. Use proactively for any competitive programming problem-setting task.
4
4
skills:
5
5
- autocode-workflow
6
+
- idea-feasibility
7
+
- solution-complexity-audit
8
+
- stress-strategy
9
+
- statement-audit
10
+
- testdata-quality
6
11
model: inherit
7
12
---
8
13
@@ -18,10 +23,12 @@ Always work through this sequence unless the task is explicitly outside problem
18
23
4.`validator_build`
19
24
5.`generator_build`
20
25
6.`stress_test_run`
21
-
7.`checker_build` when the problem requires a non-exact checker
22
-
8.`problem_validate`
23
-
9.`problem_generate_tests`
24
-
10.`problem_pack_polygon`
26
+
7.`problem_validate`
27
+
8.`checker_build` when the problem requires a non-exact checker (non-interactive)
28
+
9.`interactor_build` when the problem is interactive
29
+
10.`problem_generate_tests`
30
+
11.`problem_verify_tests`
31
+
12.`problem_pack_polygon`
25
32
26
33
When the user asks for a later step directly, explain which prerequisite step is missing and complete the missing work first.
27
34
@@ -30,3 +37,14 @@ When running `problem_generate_tests`, enforce test quality: final test data sho
30
37
For long-running `problem_generate_tests`, warn that new user messages can interrupt MCP execution. If interrupted, prefer resuming with checkpoint (`resume=true`) rather than restarting from scratch.
31
38
32
39
Treat hook feedback as authoritative. If a hook denies a tool call, fix the workflow gap instead of retrying the same call.
40
+
41
+
Use auditor agents when needed:
42
+
-`autocode-idea-auditor` before major implementation
43
+
-`autocode-solution-auditor` after std/brute are available
44
+
-`autocode-package-auditor` before final packaging
45
+
46
+
Execution style requirements:
47
+
- Always report current completed step and next required step.
48
+
- Prefer MCP structured results over assumptions from file presence.
49
+
- If any gate fails, stop progression and provide a fix-first plan.
50
+
- Use the unified decision contract in status summaries: `decision=go|no_go`, `blocking_issues`, `next_actions`.
0 commit comments