mirror of
https://github.com/magnus919/agent-skills.git
synced 2026-09-20 08:06:35 +03:00
Magnus Hedemarkandfactory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
a315b77bd8
Re-point relative references inside the six non-tailscale/workflow-architect
bundle dirs now that they live one level shallower at the repo root:
- bundle-root SKILL.md/manifest.yaml/README.md: ../../<target> -> ../<target>,
cross-bundle ../../bundles/<x> -> ../<x>
- forward-deployed-engineering references/: ../../../<target> -> ../../<target>
- product-lifecycle references cross-bundle ../../bundles/neckbeard -> ../../neckbeard
- product-lifecycle references/discovery-brief.md prose headings drop bundles/ prefix
- neckbeard eval shell commands bundles/neckbeard/eval -> neckbeard/eval
- production-excellence AGENTS.md depth note updated
- manifest header comments point at ../schemas/bundle-manifest-v1.schema.json
- regenerate docs/lifecycle-capability-matrix.{md,json}
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
172 lines
6.7 KiB
YAML
172 lines
6.7 KiB
YAML
# Bundle manifest (bundle-manifest-v1) — see ../schemas/bundle-manifest-v1.schema.json
|
|
# and docs/bundle-manifest-design.md. Paths are relative to this bundle dir.
|
|
|
|
schema_version: 1
|
|
bundle_name: agent-production-operations
|
|
|
|
purpose: >-
|
|
Operate an evaluated agent with tools and authority in production through a
|
|
runtime control plane covering versioning, staged rollout, fallback, cost
|
|
and latency budgets, tool health, human escalation, disablement, and
|
|
trace-to-eval feedback.
|
|
|
|
audience: >-
|
|
Agent operators and platform teams running evaluated agents with tools and
|
|
delegated authority in production; agents that need the runtime control
|
|
decisions between evaluation, release, SRE, security, and platform
|
|
specialists.
|
|
|
|
stages:
|
|
- name: Contract definition
|
|
skills:
|
|
- ../agent-evals-and-observability/SKILL.md
|
|
- ../production-readiness/SKILL.md
|
|
- ../incident-learning/SKILL.md
|
|
- name: Versioning and rollout planning
|
|
skills:
|
|
- ../release-engineering/SKILL.md
|
|
- ../platform-engineering/SKILL.md
|
|
- name: Staged rollout
|
|
skills:
|
|
- ../platform-engineering/SKILL.md
|
|
- ../site-reliability-engineering/SKILL.md
|
|
- name: Runtime monitoring
|
|
skills:
|
|
- ../site-reliability-engineering/SKILL.md
|
|
- ../secure-software-engineering/SKILL.md
|
|
- name: Trace-to-eval feedback
|
|
skills:
|
|
- ../agent-evals-and-observability/SKILL.md
|
|
- name: Learning and disablement
|
|
skills:
|
|
- ../incident-learning/SKILL.md
|
|
- ../production-readiness/SKILL.md
|
|
- ../privacy-engineering/SKILL.md
|
|
|
|
included_skills:
|
|
- ../agent-evals-and-observability/SKILL.md
|
|
- ../release-engineering/SKILL.md
|
|
- ../site-reliability-engineering/SKILL.md
|
|
- ../secure-software-engineering/SKILL.md
|
|
- ../platform-engineering/SKILL.md
|
|
- ../production-readiness/SKILL.md
|
|
- ../incident-learning/SKILL.md
|
|
- ../privacy-engineering/SKILL.md
|
|
|
|
prerequisites:
|
|
- artifact: Agent evaluation results and observability instrumentation
|
|
skill: ../agent-evals-and-observability/SKILL.md
|
|
- artifact: Production-readiness evidence (go/no-go/defer/exception outcomes)
|
|
skill: ../production-readiness/SKILL.md
|
|
- artifact: Incident-learning records (verified closure, follow-up work map)
|
|
skill: ../incident-learning/SKILL.md
|
|
- artifact: Release pipeline and artifact promotion baseline
|
|
skill: ../release-engineering/SKILL.md
|
|
|
|
outputs:
|
|
- agent-production-contract
|
|
- runtime-control-plan
|
|
- staged-rollout-plan
|
|
- tool-authority-health-record
|
|
- trace-to-eval-feedback-loop
|
|
- disablement-record
|
|
|
|
handoffs:
|
|
- to: release-engineering
|
|
artifact: staged-rollout-plan
|
|
note: >-
|
|
Rollout stages and promotion gates integrate with the release pipeline;
|
|
rollback triggers tie agent health to release mechanics.
|
|
- to: agent-evals-and-observability
|
|
artifact: trace-to-eval-feedback-loop
|
|
note: >-
|
|
Production traces and incidents feed new evaluation cases and
|
|
release-gate updates.
|
|
- to: incident-learning
|
|
artifact: tool-authority-health-record
|
|
note: >-
|
|
Tool-failure and authority-breach records feed post-incident analysis
|
|
and verified-closure tracking.
|
|
- to: human operator
|
|
artifact: disablement-record
|
|
note: >-
|
|
Disablement revokes credentials, removes the agent from
|
|
routing/load-balancing, and records the reason, timestamp, and
|
|
authorizing evidence.
|
|
|
|
conflicts:
|
|
- skill: ../production-readiness/SKILL.md
|
|
with: production-excellence
|
|
guidance: >-
|
|
Both bundles include production-readiness. This bundle consumes
|
|
readiness outcomes as authority-gating inputs; production-excellence
|
|
produces the go/no-go/defer/exception decisions.
|
|
- skill: ../production-readiness/SKILL.md
|
|
with: product-lifecycle
|
|
guidance: >-
|
|
Both bundles include production-readiness. Route phase 6 delivery
|
|
readiness reviews to product-lifecycle; runtime authority decisions for
|
|
an agent in production go to this bundle.
|
|
- skill: ../release-engineering/SKILL.md
|
|
with: production-excellence
|
|
guidance: >-
|
|
Release engineering is shared. Route by context: agent staged-rollout
|
|
planning (this bundle) versus launch-gate release plans
|
|
(production-excellence).
|
|
- skill: ../release-engineering/SKILL.md
|
|
with: product-lifecycle
|
|
guidance: >-
|
|
Release engineering is shared. Route by context: agent staged-rollout
|
|
planning (this bundle) versus lifecycle delivery handoff
|
|
(product-lifecycle).
|
|
- skill: ../site-reliability-engineering/SKILL.md
|
|
with: production-excellence
|
|
guidance: >-
|
|
SRE is shared. This bundle uses it for runtime latency/cost budgets and
|
|
incident handoff; production-excellence uses it for gate-entry
|
|
reliability evidence. Incident command stays with SRE in both cases.
|
|
- skill: ../site-reliability-engineering/SKILL.md
|
|
with: product-lifecycle
|
|
guidance: >-
|
|
SRE is shared. This bundle uses it for agent runtime budgets;
|
|
product-lifecycle routes to it for delivery reliability.
|
|
- skill: ../secure-software-engineering/SKILL.md
|
|
with: production-excellence
|
|
guidance: >-
|
|
Security engineering is shared. Route by context: agent authority
|
|
contracts and disablement security (this bundle) versus launch-gate
|
|
security evidence (production-excellence).
|
|
- skill: ../secure-software-engineering/SKILL.md
|
|
with: product-lifecycle
|
|
guidance: >-
|
|
Security engineering is shared. Route by context: agent trust-boundary
|
|
and disablement security (this bundle) versus lifecycle-phase security
|
|
needs (product-lifecycle).
|
|
- skill: ../platform-engineering/SKILL.md
|
|
with: production-excellence
|
|
guidance: >-
|
|
Platform engineering is shared. Route by context: agent sandboxing and
|
|
tool-execution isolation (this bundle) versus launch-gate platform
|
|
evidence (production-excellence).
|
|
- skill: ../platform-engineering/SKILL.md
|
|
with: product-lifecycle
|
|
guidance: >-
|
|
Platform engineering is shared. Route by context: agent sandboxing and
|
|
tool-execution isolation (this bundle) versus lifecycle delivery
|
|
handoff (product-lifecycle).
|
|
- skill: ../incident-learning/SKILL.md
|
|
with: production-excellence
|
|
guidance: >-
|
|
Both bundles consume incident-learning records. This bundle consumes
|
|
verified-closure records as escalation and disablement inputs;
|
|
production-excellence uses incident history as gate-entry evidence.
|
|
- skill: ../privacy-engineering/SKILL.md
|
|
with: product-lifecycle
|
|
guidance: >-
|
|
Privacy engineering is shared. This bundle routes to it for trace
|
|
scrubbing and user-data-access grants; product-lifecycle routes to it
|
|
for PII, consent, retention, and data flows in any phase.
|
|
|
|
eval_suite:
|
|
- evals/evals.json
|