Skill Discovery and Rule Management

Chooses which skills apply to a task, and says how rules are added and kept in order.

Always

Load every skill that might apply before the first answer or question, and name them: a skill loaded late cannot change the plan
A new rule goes into the one skill that owns the topic; other skills point to it as skill-name: Exact Section Title
Write only what a strong model would not do by default here: project choices, real paths and commands, and the reason for each rule
Every example follows the other skills' rules
SKILL.md stays an overview of at most 80 lines that links every reference file with a "Read when"
Run node .agents/skills/eser-rules-manager/scripts/validate-skills.ts --fix after every skill change
A change meant to alter behavior is evaluated against a snapshot of the old skill

Skill	Covers
`enrichment`	Round-based interview until no decision is left open
`handoff`	`/handoff`: a document a fresh session continues from
`workflow-practices`	Clarifying, roles, approvals, task flow, gates
`architecture-guidelines`	Hexagonal layers, API surface, testing, ADRs
`design-principles`	Pure functions, immutability, interfaces, clocks
`coding-practices`	Errors, logging, resources, config, readability
`javascript-practices`	TS/JS modules, types, async, React
`go-practices`	Go conventions, context, data structures, perf
`security-practices`	Secrets, input, SSRF, injection, LLM trust boundary
`tooling-standards`	Tooling, dependencies, scripts, naming, packages
`documentation-conventions`	READMEs, docs and prose style
`release-and-ci`	Versions, releases, publishing, workflows, deploys

File	Read when
skill-discovery.md	Unsure which skills a task needs
skill-format.md	Writing or restructuring a skill or a rule
skill-evaluation.md	Testing that a skill changes behavior and triggers