Work better with AI
Configure project instructions and reusable skills
Put stable project guidance and recurring procedures where your AI tool can discover them.
What you will work toward.
Concise project instructions and one scoped reusable skill with tool-correct placement and a discovery check.
Make this workflow yours.
Choose your AI tool, describe your task, and copy one complete project brief. Your inputs stay in this tab.
Includes your prompt, selected references, skill instructions, setup steps, and checks. Use your own AI account; Pristren does not run a model here.
Describe your project or task to prepare your brief.
Read the complete brief
Configure project instructions and reusable skills for [Describe the project, audience, and useful result.] within Use existing project conventions. State assumptions and verify the result.. Inspect existing instruction files, rules, and skill directories before adding anything. Identify the target host and its current official discovery rules. Separate stable project facts such as build commands and conventions from a recurring task procedure that should load only when relevant. Write concise project guidance in the host-supported location without contradicting more specific instructions. Create one focused SKILL.md with valid name and description frontmatter, clear invocation intent, required inputs, an execution procedure, and observable completion checks. Keep it instruction-only unless a script or asset has a concrete need. Avoid unrelated examples and duplicated global policy. Use the correct project-local skill path for the selected tool. Explain how to discover or invoke the skill and verify that the intended version loads. Run its procedure on a representative safe example only if the tool is available; otherwise label compatibility documentation-reviewed. Report files, scope, discovery evidence, and any untested behavior. Do not assume identical skill behavior across hosts. Execution environment: Codex. Follow applicable project instructions and verify the requested result. Workflow https://pristren.com/workflows/project-instructions-and-skills/ Expected result Concise project instructions and one scoped reusable skill with tool-correct placement and a discovery check. Before you begin - The target AI tool and a repository. - A real recurring task with example inputs and expected output. - Existing rules/instruction files and any team conventions. Setup for Codex - Start Codex in the repository and read applicable AGENTS.md guidance. Include only the files this task needs. - Use .agents/skills/<skill-name>/SKILL.md for this host’s project-local skill. Inspect the original resource’s full install guidance. - Inspect existing guidance first. Put stable repository facts in project instructions and task-specific procedures in a small skill. - Use the acceptance checks below to assess the result; an agent completion message is not evidence that the task passed. Selected resources - Configure project instructions and reusable skills — master prompt: https://pristren.com/resources/project-instructions-and-skills-prompt/ Use the task-specific master prompt above. - Build reports from evidence: https://pristren.com/resources/evidence-led-report/ Use the published resource’s Download SKILL.md action, or copy its complete original SKILL.md text. Draft resources are not publicly downloadable. Save the supplied SKILL.md as .agents/skills/evidence-led-report/SKILL.md in the project. Review it before use. Open the project in your tool and select evidence-led-report by name; check that it resolves to this project-local version. Read the resource below and apply only the parts relevant to the current task. Inspect the loaded skill before using it on a representative task. This page documents setup; it has not performed the installation for you. mkdir -p .agents/skills/evidence-led-report Skill instructions (Build reports from evidence) --- name: evidence-led-report description: Produce a sourced research or business report from supplied data while keeping observations, calculations, and recommendations distinct. --- # Evidence-led report Define the question, audience, time range, and unit of analysis before drawing conclusions. Identify the source of each input and whether it is current enough for the question. Ask only for missing information that would materially change the result. Keep a working evidence record: source URL or artifact, checked date, claim supported, scope or plan qualifier, and uncertainty. Missing documentation means unverified; it does not establish that a capability is absent. A vendor claim is not a hands-on benchmark. For calculations, state units, denominators, exclusions, and transformations. Check that totals reconcile and distinguish zero from missing values. Do not turn views into unique people, a download into a successful implementation, or an enquiry into a paying client. Write the report around the answer and the evidence needed to assess it. Separate observed facts, interpretations, and proposed actions. Cite close to the supported statement. Include material limitations and make recommendations proportional to the evidence. For a repeatable report, describe the expected inputs and validation checks so a later run can reject stale or malformed data. Drafting a report does not authorize sending it, publishing it, or modifying the source system. Execution 1. Inventory existing guidance and find one repeated task. 2. Separate stable repository rules from task-specific procedure. 3. Author a small skill and place it in the selected host’s project directory. 4. Check formatting, discovery, and a representative invocation when the host is available. Acceptance checks - Frontmatter includes a valid name and a useful nonempty description. - The skill has a distinct purpose and observable completion criteria. - Project guidance avoids conflicting or duplicated instructions. - Discovery and hands-on execution are reported separately. If something fails - The wrong skill version loads: Inspect project and user-level paths and identify the resolved copy before changing instructions. - The skill activates for unrelated tasks: Narrow its description and scope; move broad repository facts into project guidance. Inspect the project first. Report the changes, the checks actually run, and anything still unresolved. Ask before any deployment or external publication.
Tool setup and prerequisites
Tool setup checked against official documentation. Pristren has not executed this workflow through this tool.
- Start Codex in the repository and read applicable AGENTS.md guidance. Include only the files this task needs.
- Use .agents/skills/<skill-name>/SKILL.md for this host’s project-local skill. Inspect the original resource’s full install guidance.
- Inspect existing guidance first. Put stable repository facts in project instructions and task-specific procedures in a small skill.
- Use the acceptance checks below to assess the result; an agent completion message is not evidence that the task passed.
Before you begin
- The target AI tool and a repository.
- A real recurring task with example inputs and expected output.
- Existing rules/instruction files and any team conventions.
Skills and tools included in your brief
Required resources are included. Add optional ones when your task needs them.
Required
Start with this workflow’s original task-specific master prompt.
Review instructionsRequired
Use a small original skill as a concrete format and scope example.
Review instructionsOptional
Inspect a real instruction-only external skill and its individual license.
Review instructionsOptional
Inspect a real external skill with supporting files and host-specific execution requirements.
Review instructionsWhat your AI will work through.
- Inventory existing guidance and find one repeated task.
- Separate stable repository rules from task-specific procedure.
- Author a small skill and place it in the selected host’s project directory.
- Check formatting, discovery, and a representative invocation when the host is available.
Check the result before calling it done.
- Frontmatter includes a valid name and a useful nonempty description.
- The skill has a distinct purpose and observable completion criteria.
- Project guidance avoids conflicting or duplicated instructions.
- Discovery and hands-on execution are reported separately.
From Pristren’s own implementation
Forma: a project map and a reusable customization skill
The running Forma example has a small AGENTS.md project map and an original SKILL.md procedure. Read both in the browser and use the relevant instructions in your own project.
Open the mapped projectThe problem
An agent needs to discover the stack, run command, editable files, and checks before making a visual change.
The implementation
The project map records stable facts. The skill describes the recurring customization task, including preserving keyboard controls and verifying the result.
# Forma project map Scope: this Forma example directory. Stack: plain HTML, CSS, and browser JavaScript; no build step. Content: index.html. Layout and sculpture: styles.css. Interactions: script.js. Run: node server.mjs (http://127.0.0.1:4173). Check: node --check script.js and node --check server.mjs. Procedure: read SKILL.md and README.md before customization. Keep keyboard controls, visible focus, and static reduced motion. Keep sample projects labeled until real content is supplied. Publishing needs a separate user instruction.
What was checked and what remains unproven
- Run commands were checked against the included Node server and README.
- The customization skill was read while inspecting this example in Codex.
- The example's scripts pass Node.js syntax checks.
This records use in the current Codex task. Claude Code and Cursor discovery behavior has not been executed here; their setup guidance remains documentation reviewed.
When something gets in the way.
The wrong skill version loads
Inspect project and user-level paths and identify the resolved copy before changing instructions.
The skill activates for unrelated tasks
Narrow its description and scope; move broad repository facts into project guidance.
How did your attempt go?
Try the brief in your AI tool first. Your feedback helps us find workflows that need attention.
Go a little deeper.
Sources and review notes
- Codex skill guidanceReviewed 2026-11-17
- Claude Code skill guidanceReviewed 2026-11-17
- Cursor skill guidanceReviewed 2026-11-17