Copiez ce prompt dans l’assistant IA que vous utilisez.
Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/bstBizEra/badf.Chargement du README du dépôt…
Affichage de 40 skills collectés sur 45.
This skill should be used for advanced LLM evaluation: LLM-as-judge systems, direct scoring, pairwise comparison, rubric calibration, evaluator bias mitigation, confidence scoring, and automated quality assessment.
Langue du texte source : anglais
This skill should be used when modeling agent mental states with BDI concepts: beliefs, desires, intentions, RDF-to-belief transformations, rational agency traces, cognitive agents, BDI ontologies, and neuro-symbolic AI integration.
Langue du texte source : anglais
This skill should be used when long-running agent sessions need context compression, structured summarization, compaction, token-per-task optimization, or durable handoff summaries that preserve decisions, files, risks, and next actions.
Langue du texte source : anglais
This skill should be used for diagnosing and mitigating context degradation: lost-in-middle failures, context poisoning, context clash, context confusion, attention-pattern issues, and agent performance degradation caused by accumulated or conflicting context.
Langue du texte source : anglais
This skill should be used to explain or reason about the foundational concepts of context engineering: what context is, the anatomy of a context window, how attention mechanics work, the U-shaped attention curve, why context quality matters more than…
Langue du texte source : anglais
This skill should be used for improving context efficiency: context budgeting, observation masking, prefix or KV-cache strategy, partitioning, token-cost reduction, retrieval scoping, and extending effective context capacity without lowering answer quality.
Langue du texte source : anglais
This skill should be used when building agent evaluation systems: deterministic checks, regression suites, multi-dimensional rubrics, quality gates, production monitoring, baseline comparison, and outcome measurement for agent pipelines.
Langue du texte source : anglais
This skill should be used when agent work needs file-backed context: durable scratchpads, tool-output offloading, just-in-time discovery, cross-agent handoff files, filesystem memory, or cleanup policies for context stored outside the prompt.
Langue du texte source : anglais
This skill should be used when designing autonomous agent harnesses: research loops, evaluation scaffolds, locked and editable surfaces, durable logs, novelty gates, pruning, rollback, PR preparation, and human approval boundaries.
Langue du texte source : anglais
This skill should be used when the user asks to "share memory between agents", "KV cache compaction for multi-agent", "orchestrator worker context", "latent briefing", "reduce worker tokens", "cross-agent memory without summarization", or discusses Attention…
Langue du texte source : anglais
This skill should be used when writing, enhancing, or evaluating the launch prompt for a long-running autonomous agent or a parallel multi-agent orchestration attacking a hard problem: pseudo-formal task briefs that define terms and an exact success predicate…
Langue du texte source : anglais
This skill should be used for persistent semantic memory in agent systems: cross-session knowledge retention, entity tracking, temporal validity, graph or vector retrieval, memory consolidation, and memory benchmark selection. Route file-backed scratchpads to…
Langue du texte source : anglais
This skill should be used when designing multi-agent systems that need context isolation, supervisor or swarm coordination, explicit handoffs, parallel execution, or a decision on whether multiple agents are justified.
Langue du texte source : anglais
This skill should be used for project-level decisions about LLM-powered systems: whether an LLM is the right primitive for the task at hand, the shape of a multi-stage batch or agent pipeline, token and cost estimation, choosing between single-agent and…
Langue du texte source : anglais
This skill should be used for the tool-interface layer of an agent system specifically: writing tool descriptions agents can route on, designing tool schemas and response formats, naming conventions, actionable error recovery messages, MCP server design,…
Langue du texte source : anglais
Govern Git state, worktrees, branches, staging, commits, synchronization, private-history rewrites, pull requests, composition verification, integration, recovery, tags, and release refs for BADF work packages. Use whenever a BADF task changes Git state or…
Langue du texte source : anglais
G10 business-acceptance router. Takes the exact G09-validated candidate and records whether its observed behavior satisfies the business outcome the PRD and acceptance criteria approved — never whether a screen renders or a Playwright script completes.…
Langue du texte source : anglais
Compose the detailed specialist design contracts — UX/service behavior, authorization, data, API, accessibility — into one coherent, implementable solution for a BADF work package, and reconcile the seams between them. Use when a requirement (G02) must become…
Langue du texte source : anglais
G10 readiness aggregator. Resolves upstream gate evidence against the exact immutable release candidate and evaluates twelve readiness dimensions, then packages a readiness dossier whose strongest positive conclusion is READY_FOR_AUTHORITY. It produces the…
Langue du texte source : anglais
G09 independent pre-release validation orchestrator. Routes a G08-verified candidate through independent quality, security, performance and resilience validation classes, binds each to the exact candidate under pre-existing thresholds and an out-of-agent…
Langue du texte source : anglais
Execute exactly one authorized BADF Governed Work Package into G07 build evidence — claim, preflight, isolate, baseline, slice, test-first where a durable seam exists, implement, verify freshly, self-review, reconcile, package, hand off — inside the WP's…
Langue du texte source : anglais
Independently verify one built BADF Work Package into G08 evidence — bind the exact candidate and composed target, seal the inputs, run isolated defect-first reviews (the Reviewer plane), let an approved runtime observe integration, contract and composed-tree…
Langue du texte source : anglais
Compose an approved BADF design (G01–G05) into a governed Work Package DAG — bounded, authorized, independently verifiable execution units — and normalize the plan into the existing G06 evidence (work-breakdown, test-plan, release-plan, rollback-plan). Use…
Langue du texte source : anglais
Compose the pre-implementation security-design specialist contracts — threat model, security requirements, privacy, abuse cases, API/IAM/supply-chain security design, and (conditionally) AI-agent security — over the architecture and solution baselines, and…
Langue du texte source : anglais
Decide and bind a BADF release the way the release contract requires -- the version is a governed SemVer decision by the release authority, the binding is written with `badf_gate.py git-release-record` before the tag is created, the tag is annotated and…
Langue du texte source : anglais
Recover Git state the way badf-git's recovery contract requires -- PRESERVE, IDENTIFY and CLASSIFY before anything destructive -- by rendering the unique-state inventory with the read-only `badf_gate.py git-recovery` inspector, establishing preservation with…
Langue du texte source : anglais
Integrate a BADF work package through its pull request the way the contract requires -- re-check the current target, the exact reviewed head, the merge method, the required checks on that head and the composition claim immediately before merging; merge only…
Langue du texte source : anglais
Bind and verify what protected integration is expected to produce -- write the git-composition record with `badf_compose.py --record`, commit it inside the pull request under work/<WP>/evidence/G07/composition-record.json, and let compose recompute and…
Langue du texte source : anglais
Run badf-git's nested Git loop with intentional staging, inspected diffs and atomic commits, and keep the evidence honest across it -- bind a git-baseline record before verification and judge it with the read-only `badf_gate.py git-staleness` inspector before…
Langue du texte source : anglais
Observe and record the Git baseline of a working tree before any material Git operation -- repository identity, worktree/branch/index state as counts, target and source revisions and trees, merge base, ahead/behind, remote freshness and policy epoch -- by…
Langue du texte source : anglais
Govern architecture design, architectural decisions, architecture fitness and drift assurance for BADF-controlled delivery. Use when a system's structure, boundaries, decisions, NFR allocation or operability must be designed (DESIGN), or when an…
Langue du texte source : anglais
Decompose an approved PRD baseline into the four canonical BADF Gate G02 evidence artifacts — requirements, quantified NFRs, a bidirectional traceability matrix (RTM), and a definition-of-ready packet. Use when converting product objectives into testable,…
Langue du texte source : anglais
Govern evidence-based research for BADF questions, issues, defects, architecture uncertainties, solution alternatives, repositories, technologies, standards and claims. Use when material uncertainty must be reduced before a BADF decision, work package,…
Langue du texte source : anglais
Answer an empirical question by running a controlled experiment, not by citing sources -- state a falsifiable hypothesis, design a method that could refute it, run it under the BADF experiment mechanism, and record method and result. For research type R08…
Langue du texte source : anglais
Verify one or more explicit factual claims against evidence already bound to the research record, preserving uncertainty, contradiction and provenance. A cross-cutting primitive for research type R10 and for every multi-source route. It may strengthen or…
Langue du texte source : anglais
The terminal research subskill -- decide whether the evidence is enough to support a governed decision, and return a controlled disposition. Its question is not "what is our recommendation?" but "do we know enough?". It never returns PASS/FAIL, never…
Langue du texte source : anglais
Independently challenge a research recommendation -- attempt to refute it, find omitted or non-independent evidence, counterexamples, unsupported assumptions and uninspected surface. It is separate from the originating research run and casts sealed ballots…
Langue du texte source : anglais
Evaluate the candidate alternatives against explicit, comparable criteria and produce an evidence-backed trade-off -- surfacing non-dominance rather than manufacturing a winner. For research type R07 (comparative). It compares; it does not gather evidence…
Langue du texte source : anglais
Find the technically viable approaches that could solve a problem -- an option set, not one early answer. For research types R04 (technical solution) and R05 (architecture). Produces candidate solutions grounded in the record's evidence, with their…
Langue du texte source : anglais
Acquire external evidence for a research question -- decompose the question, discover and prefer authoritative/primary sources, corroborate across independent sources, capture provenance, and preserve contradictory evidence. Read-only acquisition. It does not…
Langue du texte source : anglais
Fichier source du README : README.md
Lire le README source