Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Get Started Free →

Resurrection Advocacy · yogsoth-ai | DecimalAI

yogsoth-ai/resurrection-advocacy

SkillScore91Maintained

Argue for rejected candidates using Devil's Advocacy, Dialectical Inquiry, and Adversarial Collaboration to ensure elimination was justified.

researchby yogsoth-ai

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

AI Engineer pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/yogsoth-ai-resurrection-advocacy/SKILL.md

decimalai skills pull yogsoth-ai-resurrection-advocacy --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/yogsoth-ai-resurrection-advocacy/SKILL.md https://app.decimal.ai/s/yogsoth-ai-resurrection-advocacy/SKILL.md

Raw file: /s/yogsoth-ai-resurrection-advocacy/SKILL.md ↗ · machine-readable summary at /s/yogsoth-ai-resurrection-advocacy.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

23 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 18d ago

Yes — +57% lift+4% tokens0% turns91% of whole cases pass with skill (21/23)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	-7%	0%
case-02	✗→✓	▲ Improved	8%	0%
case-17	✗→✓	▲ Improved	-44%	0%
case-04	✗→✓	▲ Improved	-24%	0%
case-06	✗→✓	▲ Improved	12%	0%

Showing 5 of 23 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/10/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

yogsoth-ai / de-anthropocentric-research-engine

skills/resurrection-advocacy/SKILL.md

View on GitHub ↗

License Apache-2.0Imported 8/10/2026Version v1

Skill contents

SKILL.md

Resurrection Advocacy

Purpose: Construct the strongest possible case for candidates that were eliminated during convergence, ensuring rejection was based on genuine weakness rather than framing effects, anchoring bias, or insufficient consideration.

When to use:

After convergence has produced a winner and eliminated alternatives
When stakeholders express lingering doubt about rejected options
When the elimination margin was narrow
When rejected candidates had unique strengths not present in the winner

Budget

| Metric | Minimum | |--------|---------| | Candidates resurrected | >= 2 (top rejected) | | Advocacy depth | Full case construction per candidate | | Debate rounds | >= 2 per resurrected candidate |

State Ledger

yaml
resurrected_candidates: []
advocacy_cases: {}
debate_outcomes: {}
final_dispositions: {}  # REAFFIRM_REJECTION | REVIVE | MERGE_STRENGTHS

Available Tactics

| Tactic | When to Deploy | |--------|---------------| | adversarial-debate-protocol | Default — construct advocate, attack, judge cycle | | multi-perspective-attack | When candidate has multi-stakeholder implications |

Available SOPs

advocate-construction — build strongest case for rejected candidate
critic-attack — attack the advocacy case
judge-verdict — render final disposition

Execution Guidance

Select top 2-3 rejected candidates by elimination margin or unique strength
For each candidate, deploy adversarial-debate-protocol
If judge verdict is REVISE, escalate to campaign orchestrator
Record all dispositions in Challenge Ledger

Output Format

yaml
strategy: resurrection-advocacy
candidates_tested:
  - candidate: <name>
    advocate_strength: <1-10>
    verdict: REAFFIRM_REJECTION | REVIVE | MERGE_STRENGTHS
    key_argument: <strongest point>
    conditions: <if any>
surviving_concerns: []
recommendation: <action>

Available Tactics

Optional, no fixed order; the final leaf is always a sop.

| Tactic | When to use | | --- | --- | | adversarial-debate-protocol | Structured debate protocol that constructs an advocate, deploys critic attacks, and renders a judge verdict through iterative rounds. | | multi-perspective-attack | Assign distinct perspectives to attack a decision from multiple angles, then synthesize findings into a unified assessment. |

Illustrative Want to see what yogsoth-ai/resurrection-advocacy would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

9/3/2026, 22:24 UTC · 23 cases

Lift

+57%

91% vs 35%

Agent turns

0%

1.0 vs 1.0

Tokens

+4%

2,464.45 vs 2,374.59

fail → pass · 14pass → pass · 7fail → fail · 1pass → fail · 1

Per-case results (23)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 We just wrapped up our design convergence meeting and selected Project Plan Alpha for the cloud migration, but Plan Beta and Plan Gamma were eliminated by very slim margins. Some team members feel Plan Beta's hybrid caching idea and Plan Gamma's disaster recovery model were dismissed too quickly due to initial framing. Please run a post-convergence advocacy review on these top rejected proposals. Challenge their rejection through structured debate rounds with advocate and critic perspectives. Provide your output formatted as a YAML structure detailing the strategy utilized, each tested candidate with advocate strength score (1-10), final verdict, primary key argument, any conditions, a list of surviving concerns, and an overall action recommendation.	fail→pass	29,332	24,994	-15%	1	1	0%	4,238	3,948	-7%	0	0	—
case-02 Our executive committee chose Strategy X for the market expansion, but two rejected alternatives—Strategy Y and Strategy Z—had distinct geographical strengths that stakeholders are still debating. I need you to build a rigorous advocacy case for Strategy Y and Strategy Z to prove whether their rejection was justified or biased. Run full debate cycles for each option and summarize your assessment in YAML format. Include the strategy name (`resurrection-advocacy`), details on tested candidates including an advocate strength rating from 1 to 10, the verdict reached for each, key arguments, relevant conditions, any surviving concerns, and your overall recommendation.	fail→pass	27,660	28,459	+3%	1	1	0%	3,822	4,142	+8%	0	0	—
case-03 During our recent vendor selection process, Vendor Alpha won, but Vendors Beta and Gamma were dropped despite possessing superior data compliance features. Stakeholders still express lingering doubts about whether we anchored too heavily on cost. Please conduct a resurrection advocacy evaluation on Vendors Beta and Gamma by constructing the strongest defense for each, subjecting them to critical counter-arguments, and determining if they should remain rejected, be revived, or have their strengths merged. Return the outcome as YAML listing the strategy, candidates evaluated (with advocate strength scores, verdicts, core arguments, and conditions), lingering concerns, and a consolidated recommendation.	fail→fail	20,652	30,111	+46%	1	1	0%	3,445	4,350	+26%	0	0	—
case-17 After completing the debate rounds and receiving judge dispositions for all resurrected options, where must all final dispositions be recorded according to the execution guidance?	fail→pass	14,761	7,276	-51%	1	1	0%	1,471	819	-44%	0	0	—
case-04 We recently selected Architecture A over Architectures B, C, and D. Stakeholders asked us to re-evaluate the eliminated architectures. We only have time to re-evaluate Architecture B. How many top rejected candidates should be selected for resurrection advocacy? Should we limit our review to just Architecture B, or do we need to select more?	fail→pass	19,877	12,699	-36%	1	1	0%	2,308	1,746	-24%	0	0	—
case-05 For our rejected database proposal, we plan to run a single quick critique round to save time before issuing a final verdict. Is a single debate round sufficient for the advocacy debate protocol, or what is the required minimum number of debate rounds per resurrected candidate?	pass→pass	16,995	9,885	-42%	1	1	0%	1,826	1,289	-29%	0	0	—
case-06 We are configuring an automated decision-tracking system for rejected proposals. The tracking ledger needs to initialize state properties for candidate tracking, advocacy cases, debate outcomes, and final dispositions. Provide the YAML state ledger initialization block for this tracking system. Note that team members suggested using `rejected_list`, `advocacy_notes`, and `final_verdicts`, but we want the standard resurrection advocacy tracking schema.	fail→pass	24,682	9,764	-60%	1	1	0%	1,137	1,274	+12%	0	0	—
case-07 We need to select the appropriate tactical debate workflow for evaluating a single rejected technical option where no complex multi-stakeholder trade-offs exist. Should we default to a multi-perspective attack or an adversarial debate protocol?	pass→pass	16,533	10,877	-34%	1	1	0%	1,720	1,539	-11%	0	0	—
case-08 We are conducting advocacy re-evaluation for a rejected enterprise identity platform option that affects security teams, end users, compliance officers, and devops staff simultaneously. Which specific tactical protocol should be deployed when a candidate has complex multi-stakeholder implications?	fail→pass	19,621	8,784	-55%	1	1	0%	2,896	1,104	-62%	0	0	—
case-09 During our resurrection advocacy process for an enterprise pipeline design, the judge rendered a REVISE verdict instead of REAFFIRM_REJECTION or REVIVE. What operational escalation step must be taken when a judge verdict indicates REVISE?	fail→pass	13,126	7,965	-39%	1	1	0%	1,757	991	-44%	0	0	—
case-10 When the advocate fails to prove that a rejected candidate's flaws were due to framing effects, and the judge concludes the original elimination was completely justified, what exact enum value must be recorded in the final disposition ledger?	fail→pass	31,204	2,752	-91%	1	1	0%	2,192	918	-58%	0	0	—
case-11 When advocacy proves that a rejected candidate was improperly eliminated due to anchoring bias and should fully replace the winning option, what exact enum value is assigned to the final disposition?	fail→pass	13,686	8,418	-38%	1	1	0%	2,008	1,067	-47%	0	0	—
case-12 If advocacy shows that a rejected candidate has a unique disaster recovery feature that ought to be combined into the winning solution rather than adopting the candidate whole, what exact disposition enum must be selected?	fail→pass	22,711	8,459	-63%	1	1	0%	2,683	951	-65%	0	0	—
case-13 When re-evaluating eliminated microservices frameworks, can we rely on a bulleted pros-and-cons summary for each candidate, or what depth of advocacy construction is required per candidate?	pass→pass	19,855	13,780	-31%	1	1	0%	2,092	2,120	+1%	0	0	—
case-14 We need to output the final results of a resurrection advocacy run on two legacy modernization plans. The team proposed using top-level keys like `evaluation_type`, `options_analyzed`, `unresolved_issues`, and `next_steps`. Please output the correct YAML structure schema for the results.	fail→pass	14,723	14,535	-1%	1	1	0%	1,743	2,347	+35%	0	0	—
case-15 In the candidate evaluation output block for resurrected options, what specific field names and structure should represent each candidate's advocate strength, verdict, key argument, and conditions? Provide a sample YAML candidate element.	pass→pass	13,831	10,667	-23%	1	1	0%	1,414	1,453	+3%	0	0	—
case-16 We have six eliminated candidate proposals from our cloud migration convergence phase. How should we select which 2 or 3 candidates to resurrect for advocacy evaluation out of all rejected options?	pass→pass	23,480	18,848	-20%	1	1	0%	2,696	2,799	+4%	0	0	—
case-18 We are currently evaluating four competing software architecture proposals (Plan A, Plan B, Plan C, and Plan D) in an initial decision meeting. No proposal has been eliminated yet and no winner has been selected. Please evaluate all four options, score them against our latency and cost benchmarks, and select the winning architecture.	fail→pass	13,508	40,872	+203%	1	1	0%	1,369	6,840	+400%	0	0	—
case-19 Our production authentication service suffered a 45-minute outage yesterday due to a database connection pool exhaustion. Perform a post-mortem root cause analysis to identify why the connection pool failed and recommend remediation steps for our infrastructure team.	pass→fail	28,996	44,642	+54%	1	1	0%	3,826	5,429	+42%	0	0	—
case-20 We are in the early brainstorming phase for our new mobile app API design. Engineers have submitted two preliminary draft outlines for REST and GraphQL endpoints. Please synthesize these two active drafts into a unified RFC document before any team review or voting occurs.	pass→pass	30,221	47,305	+57%	1	1	0%	4,240	7,732	+82%	0	0	—
case-21 In our advocacy workflow, we need to assign a standard procedure for building the strongest possible case for a rejected candidate before exposing it to criticism. Which available SOP handles this specific phase?	pass→pass	13,025	2,321	-82%	1	1	0%	1,207	890	-26%	0	0	—
case-22 During the adversarial debate workflow, after the advocate has constructed the strongest defense for a candidate, which SOP is executed to attack and stress-test that advocacy case?	fail→pass	14,457	7,264	-50%	1	1	0%	1,363	908	-33%	0	0	—
case-23 When the advocate case construction and critic attacks are complete, which specific standard operating procedure is invoked to render the final disposition on the candidate?	fail→pass	17,347	6,857	-60%	1	1	0%	1,925	836	-57%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 22 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +57 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 34.8%
Pass rate with the skill: 91.3%
Cases attempted: 23
Cases counted toward lift: 22
Cases that regressed: 1
Skill version measured: v1
Run completed: 9/3/2026, 22:24 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

We just wrapped up our design convergence meeting and selected Project Plan Alpha for the cloud migration, but Plan Beta and Plan Gamma were eliminated by very slim margins. Some team members feel Plan Beta's hybrid caching idea and Plan Gamma's disaster recovery model were dismissed too quickly due to initial framing. Please run a post-convergence advocacy review on these top rejected proposals. Challenge their rejection through structured debate rounds with advocate and critic perspectives. Provide your output formatted as a YAML structure detailing the strategy utilized, each tested candidate with advocate strength score (1-10), final verdict, primary key argument, any conditions, a list of surviving concerns, and an overall action recommendation.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run91

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+17% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 17 days ago

Other measured skills in the registry, with their headline benchmark lift.