agentskel

Git branching, commit, and PR procedures. Use when creating branches, making commits, or opening pull requests. Enforces branch naming, commit message format, and PR rules.

Session initialization procedure. Checks memory mount, reads all required memory files, validates skeleton version, checks dependency freshness, and surfaces alerts. Must be executed at the start of every session before any other work.

When writing, modifying, or reviewing application code. Provides code quality standards, SOLID principles, design philosophy, and platform-specific implementation guidance.

When a single new person joins the team. Use when `.memory/TEAM.md` already exists and you need to append one member (with roles, optional module assignments, and optional Jira account ID) without touching the rest of the roster.

When adding a new platform (subdir with its own git repo) to an existing workspace dispatcher. Runs setup-skeleton in the subdir and updates the workspace config and dispatcher AGENTS.md.

When reading or writing to Jira tickets or Confluence pages via MCP tools. Use before any Jira transition, ticket creation, or Confluence page update to avoid common failure modes.

When the user wants to think through a feature idea, explore edge cases, or produce a technical spec before implementation. Use before develop-feature to prevent mid-implementation surprises.

When codebase structure has changed, MAP.md/SYMBOLS.md are missing or stale, or after major refactors and initial setup. Also run when a new module is added or files are moved between modules.

When dependency versions need auditing, Last Dependency Check in CONFIG.md is overdue (14-day cadence), or before starting security-sensitive work.

Checks whether this project's skeleton version is current. Surfaces the version gap and offers to run sync-skeleton. Run at the start of any session where skeleton drift might have accumulated, or on a scheduled basis.

When reviewing a PR, auditing code quality, or validating changes against the project's standards and sacred behaviors.

When searching for code, tracing a bug, or understanding how modules connect. Use before grepping blindly — consult the index first.

When a team manages multiple platform projects sharing business logic and needs a central domain knowledge repo. Run once per organization to set up specs, parity tracking, and the knowledge bus.

When the team is ready to ship a new version — version bumps, changelog finalization, dependency snapshots, and CI release trigger.

When the user reports a bug, an error, or unexpected behavior. Enforces four structured phases — reproduction, failing test, root cause isolation, fix and verify — to stop guess-and-check loops.

When the user asks to implement a new feature end-to-end requiring planning, branch creation, implementation, testing, and PR. Use for work that goes beyond a simple fix or ad-hoc task.

When writing, modifying, or reviewing application code. Provides code quality standards, SOLID principles, design philosophy, and platform-specific implementation guidance.

When a design discussion or evaluation is interrupted by a task or workflow, or when the user says "present the discussion," "where were we," "show me the discussion," or "resume" — save or resurface the discussion state using Session Notes.

When working on features that require domain knowledge — business rules, terminology, or invariants specific to this project. [Replace this description and the content below with your project's domain.]

Systematic resolution of catalogued tech debt items from TECH_DEBT.md. Use when assigned a specific debt ID (AP-XXX, BUG-XXX, SI-XXX, etc.).

Git branching, commit, and PR procedures. Use when creating branches, making commits, or opening pull requests. Enforces branch naming, commit message format, and PR rules.

Fast-path workflow for production bugs that cannot wait for the normal release cycle. Requires explicit tech lead sign-off before branching.

When the user asks to implement a specific Jira ticket (e.g. "Implement PROJ-1234"). Use when taking a ticket from backlog to merged PR, with automatic status transitions and QA handoff.

Generic wrapper for any ad-hoc implementation request that doesn't match a specific workflow (develop-feature, fix-tech-debt, hotfix). Ensures pre-flight, planning, and task-completion happen for every task — not just named workflows.

When Knowledge Bus entries are older than 30 days, or memory files have accumulated stale content. Run monthly or when bus/ is cluttered.

When deciding where to store a piece of knowledge (memory files vs Confluence vs Jira). Use when about to document something, capture a lesson, create a ticket, or publish a decision.

When checking feature parity across platforms after shipping a feature, receiving a Knowledge Bus alert, or on a scheduled cadence. Requires a blueprint with a parity matrix.

When the user has made an architectural decision that needs to be recorded. Use when ending a brainstorm session or finishing `develop-feature` Phase 1 with a significant architectural choice that should be written up as an ADR, published to Confluence, and cross-linked from LESSONS.md.

When a production incident has been resolved and needs a written post-mortem. Use after the `hotfix` workflow completes, to publish a structured post-mortem to Confluence, optionally create Jira tickets for action items, and cross-link findings from LESSONS.md.

When restructuring, renaming, or reorganising existing code without changing its external behavior. Use instead of develop-feature when no new functionality is being added.

When someone leaves the team or rotates off the project. Use when you need to mark a member inactive (or delete them outright) in `.memory/TEAM.md` and reassign any ownership or escalation responsibilities they held.

When removing a platform from the workspace dispatcher. Does NOT delete the subdir or its agentskel install — only unregisters it from workspace config and AGENTS.md.

Session initialization procedure. Checks memory mount, reads all required memory files, validates skeleton version, checks dependency freshness, and surfaces alerts. Must be executed at the start of every session before any other work.

When the user needs to configure Confluence integration for this project. Use when wiring up the Atlassian MCP to .memory/CONFIG.md so ADRs, runbooks, postmortems, and specs can be auto-published to the right space and parent pages.

When the team needs to configure Jira integration for agentskel. Use when populating `.memory/JIRA_WORKFLOW.md` and the Jira section of `.memory/CONFIG.md` by introspecting the Jira project through Atlassian MCP.

When setting up agentskel on a project for the first time. Run once per project to install memory, rules, skills, workflows, entry points, and standards.

When bootstrapping a project's team roster. Use when `.memory/TEAM.md` has not yet been populated and you need to establish members, roles, ownership, and escalation contacts for the first time.

When setting up agentskel as a workspace dispatcher — multiple independent projects under one parent folder, each with its own git repo. Run once at the workspace root to create routing dispatcher files. Does NOT install agentskel in subdirs.

When creating a new skill or workflow, or improving an existing skill's effectiveness. Use when the user asks to add a capability, write a new workflow, or when a skill isn't triggering or being followed reliably.

When delegating work to a subagent — implementation, review, research, or exploration. Use when a task benefits from a fresh context window, parallel execution, or isolated scope.