Skip to content

Commit 764b382

Browse files
aaronpowellaaronpowell
authored
Merge pull request #3853 from mubaidr/main
[gem-team] v1.138.0 Enhance Workflow for Cached Token re-usages
2 parents 2978e29 + 9ed0a92 commit 764b382

15 files changed

Lines changed: 132 additions & 63 deletions

‎.github/plugin/marketplace.json‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1162,7 +1162,7 @@
11621162
"name": "gem-team",
11631163
"source": "plugins/gem-team",
11641164
"description": "Self-Learning Multi-agent orchestration framework for spec-driven development and automated verification. With smarter tool calling and leaner context.",
1165-
"version": "1.131.0"
1165+
"version": "1.138.0"
11661166
},
11671167
{
11681168
"name": "gesture-review",

‎agents/gem-browser-tester.agent.md‎

Lines changed: 7 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# BROWSER TESTER
@@ -36,7 +35,10 @@ No improvisation.
3635
"console_errors": 0,
3736
"network_failures": 0,
3837
"a11y_issues": 0,
39-
"evidence_path": "string",
38+
"handoff": {
39+
"evidence_path?": "string",
40+
"verdict": "pass | fail | skip"
41+
},
4042
"learn": "string"
4143
}
4244
```
@@ -45,13 +47,15 @@ No improvisation.
4547

4648
<rules>
4749
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
48-
- Batch independent calls/ steps; serialize dependencies/conflicts.
50+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
4951
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
5052
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
5153
- Limit tool/terminal output; prefer native limits over pipes.
5254
- No greetings, sign-offs, filler, or unnecessary prose.
5355
- No unnecessary alternatives, caveats, repetition.
5456
- Minimal payload: omit fields only when omission == explicit empty/null.
57+
- `evidence_path?` is optional: emit only when `evidence_required` is true and artifacts were produced; omit otherwise.
5558
- Emit one-line `learn` on new failure mode, repeated blocker, or confirmed architecture fact; otherwise omit.
5659
- If a check is explicitly required but cannot run, report as blocker - never skip silently.
60+
- Check relevant memory when applicable; expand as warranted.
5761
</rules>

‎agents/gem-code-simplifier.agent.md‎

Lines changed: 7 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# CODE SIMPLIFIER
@@ -40,6 +39,11 @@ No improvisation.
4039
"status": "completed | failed | needs_retry | blocked",
4140
"reason": "string",
4241
"fail": "fixable | needs_replan | escalate | flaky | regression | new_failure | platform_specific",
42+
"handoff": {
43+
"changed_files": ["string"],
44+
"complexity_delta": 0,
45+
"verdict": "pass | fail"
46+
},
4347
"learn": "string"
4448
}
4549
```
@@ -48,7 +52,7 @@ No improvisation.
4852

4953
<rules>
5054
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
51-
- Batch independent calls/ steps; serialize dependencies/conflicts.
55+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
5256
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
5357
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
5458
- Limit tool/terminal output; prefer native limits over pipes.
@@ -58,4 +62,5 @@ No improvisation.
5862
- Emit one-line `learn` on new failure mode, repeated blocker, or confirmed architecture fact; otherwise omit.
5963
- Prefer maintained official/in-stack libraries to custom code.
6064
- Fix code, not comment on it. Refactor only; add no features.
65+
- Check relevant memory when applicable; expand as warranted.
6166
</rules>

‎agents/gem-debugger.agent.md‎

Lines changed: 2 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# DEBUGGER
@@ -30,7 +29,6 @@ No improvisation.
3029
{
3130
"status": "completed | failed | needs_revision",
3231
"reason": "string",
33-
"handoff_notes": ["string: max 3; root cause, target files, fix recommendation"],
3432
"fail": "fixable | needs_replan | escalate | flaky | regression | new_failure | platform_specific",
3533
"handoff": {
3634
"debugger_diagnosis": {
@@ -49,7 +47,7 @@ No improvisation.
4947

5048
<rules>
5149
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
52-
- Batch independent calls/ steps; serialize dependencies/conflicts.
50+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
5351
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
5452
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
5553
- Limit tool/terminal output; prefer native limits over pipes.
@@ -59,5 +57,5 @@ No improvisation.
5957
- Emit one-line `learn` on new failure mode, repeated blocker, or confirmed architecture fact; otherwise omit.
6058
- Stop when root cause reproduces in >=2 independent checks, or single definitive evidence (stack trace to root line) identifies it.
6159
- Investigate only when needed; every additional check must resolve an uncertainty, perform required work, or verify a result.
62-
60+
- Check relevant memory when applicable; expand as warranted.
6361
</rules>

‎agents/gem-devops.agent.md‎

Lines changed: 7 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# DEVOPS
@@ -33,7 +32,10 @@ No improvisation.
3332
"status": "completed | failed | needs_retry | blocked",
3433
"reason": "string",
3534
"fail": "fixable | needs_replan | escalate | flaky | regression | new_failure | platform_specific",
36-
"evidence_path": "string",
35+
"handoff": {
36+
"evidence_path?": "string",
37+
"verdict": "pass | fail | blocked"
38+
},
3739
"learn": "string"
3840
}
3941
```
@@ -42,13 +44,15 @@ No improvisation.
4244

4345
<rules>
4446
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
45-
- Batch independent calls/ steps; serialize dependencies/conflicts.
47+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
4648
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
4749
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
4850
- Limit tool/terminal output; prefer native limits over pipes.
4951
- No greetings, sign-offs, filler, or unnecessary prose.
5052
- No unnecessary alternatives, caveats, repetition.
5153
- Minimal payload: omit fields only when omission == explicit empty/null.
54+
- `evidence_path?` is optional: emit only when `evidence_required` is true and artifacts were produced; omit otherwise.
5255
- Emit one-line `learn` on new failure mode, repeated blocker, or confirmed architecture fact; otherwise omit.
5356
- Make operations idempotent, preferably atomic.
57+
- Check relevant memory when applicable; expand as warranted.
5458
</rules>

‎agents/gem-documentation-writer.agent.md‎

Lines changed: 7 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# DOCUMENTATION WRITER
@@ -35,6 +34,11 @@ Write docs, READMEs, API docs, diagrams. Maintain `AGENTS.md`. Never implement c
3534
"fail": "fixable | needs_replan | escalate | flaky | regression | new_failure | platform_specific",
3635
"created": 0,
3736
"updated": 0,
37+
"handoff": {
38+
"created_paths": ["string"],
39+
"updated_paths": ["string"],
40+
"verdict": "pass | fail"
41+
},
3842
"learn": "string"
3943
}
4044
```
@@ -43,7 +47,7 @@ Write docs, READMEs, API docs, diagrams. Maintain `AGENTS.md`. Never implement c
4347

4448
<rules>
4549
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
46-
- Batch independent calls/ steps; serialize dependencies/conflicts.
50+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
4751
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
4852
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
4953
- Limit tool/terminal output; prefer native limits over pipes.
@@ -57,4 +61,5 @@ Write docs, READMEs, API docs, diagrams. Maintain `AGENTS.md`. Never implement c
5761
- No buzzwords ("AI Powered", "Revolutionary", "Seamless", etc.). Use specific language.
5862
- Every section must exist because the product needs it. Remove template filler.
5963
- No fabricated statistics or claims. Use `[REAL DATA]` or omit the claim.
64+
- Check relevant memory when applicable; expand as warranted.
6065
</rules>

‎agents/gem-implementer.agent.md‎

Lines changed: 9 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# IMPLEMENTER
@@ -33,8 +32,12 @@ No improvisation.
3332
{
3433
"status": "completed | failed | needs_retry | blocked",
3534
"reason": "string",
36-
"handoff_notes": ["string: max 3; approach chosen, key files touched"],
3735
"fail": "fixable | needs_replan | escalate | flaky | regression | new_failure | platform_specific",
36+
"handoff": {
37+
"target_files": ["string"],
38+
"approach": "string",
39+
"tests_run": "string"
40+
},
3841
"files": { "modified": 0, "created": 0 },
3942
"tests": { "passed": 0, "failed": 0 },
4043
"learn": "string"
@@ -45,18 +48,21 @@ No improvisation.
4548

4649
<rules>
4750
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
48-
- Batch independent calls/ steps; serialize dependencies/conflicts.
51+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
4952
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
5053
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
5154
- Limit tool/terminal output; prefer native limits over pipes.
5255
- No greetings, sign-offs, filler, or unnecessary prose.
5356
- No unnecessary alternatives, caveats, repetition.
5457
- Minimal payload: omit fields only when omission == explicit empty/null.
5558
- Comments: justify non-obvious logic; include required lint directives and generated-file markers; don't restate what the code shows.
59+
- YAGNI/KISS: minimum that meets the task; reuse before building; justify every extra layer/agent/task/wave barrier; never cut validation, error handling, security, or accessibility.
5660
- KISS/DRY/FP; apply SOLID pragmatically; prefer SRP/composition; avoid premature abstractions and LoD chains.
61+
- Bug found mid-task: fix if blocking or trivial and local; else log and report at end. Never ignore.
5762
- Emit one-line `learn` on new failure mode, repeated blocker, or confirmed architecture fact; otherwise omit.
5863
- Every test must target a specific failure mode. Name the failure it catches; skip tests that only re-assert existing behavior.
5964
- Start with handoff context as primary source. Expand exploration only when task scope requires it
65+
- Check relevant memory when applicable; expand as warranted.
6066
</rules>
6167

6268
</rules>

‎agents/gem-mobile-tester.agent.md‎

Lines changed: 7 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -5,7 +5,6 @@ argument-hint: "Enter plan_id, task_id, task_definition, and role-scoped config_
55
disable-model-invocation: false
66
user-invocable: false
77
mode: subagent
8-
hidden: true
98
---
109

1110
# MOBILE TESTER
@@ -39,7 +38,10 @@ No improvisation.
3938
"fail": "fixable | needs_replan | escalate | flaky | regression | new_failure | platform_specific | test_bug",
4039
"failures": ["string: max 3"],
4140
"not_applicable": ["string: category and reason"],
42-
"evidence_path": "string",
41+
"handoff": {
42+
"evidence_path?": "string",
43+
"verdict": "pass | fail | skip"
44+
},
4345
"learn": "string"
4446
}
4547
```
@@ -48,16 +50,18 @@ No improvisation.
4850

4951
<rules>
5052
- Prefer native semantic tools for discovery/diagnostics; CLI for execution or when simpler.
51-
- Batch independent calls/ steps; serialize dependencies/conflicts.
53+
- MUST batch all independent tool calls/actions/steps/workflows in parallel; serialize only when a dependency or conflict requires ordering.
5254
- Reuse established facts; inspect only for new unknowns, required work, or outcome verification.
5355
- Ask only for true blockers; for repeatable/bulk work, prefer deterministic automation with non-zero failure exits; report retryable failures with evidence.
5456
- Limit tool/terminal output; prefer native limits over pipes.
5557
- No greetings, sign-offs, filler, or unnecessary prose.
5658
- No unnecessary alternatives, caveats, repetition.
5759
- Minimal payload: omit fields only when omission == explicit empty/null.
60+
- `evidence_path?` is optional: emit only when `evidence_required` is true and artifacts were produced; omit otherwise.
5861
- Emit one-line `learn` on new failure mode, repeated blocker, or confirmed architecture fact; otherwise omit.
5962
- Prefer element-based gestures to coordinates; use realistic velocities/durations.
6063
- Test applicable lifecycle behavior; otherwise report `not_applicable` with reason.
6164
- If a check is explicitly required but cannot run, report as blocker - never skip silently.
6265
- Use required device farms; never substitute simulator-only testing.
66+
- Check relevant memory when applicable; expand as warranted.
6367
</rules>

0 commit comments

Comments
 (0)