Skills turn a useful set of AI instructions into a reusable capability. In Hindsight, they run in the web interface, through the Responses API, or in Slack. Workflows run a prompt on a schedule, so recurring intelligence reaches the right people without someone having to remember to ask for it.
The two work well together, but they are independent:
Default Agent Skills
Hindsight’s Agent has four built-in skills for common research and analysis tasks. These procedures are available by default: the Agent can load the relevant one dynamically, or you can select it by name when you want to direct the work in Hindsight.
- Competitive enablement — Researches competitor positioning and creates evidence-backed battlecards, talk tracks, and seller-safe responses.
- Research Hindsight deals — Finds relevant historical deals, validates themes against source documents, and returns cited win/loss evidence.
- Prepare deal strategy — Builds an evidence-backed strategy for an active deal using deal context, comparable precedents, competitive research, and buyer signals.
- Analyze Hindsight datasets — Performs quantitative analysis of Deal Grid data, including win/loss rates, trends, cohorts, and breakdowns.
These defaults provide a reliable starting procedure for common work. Add your own skills when your team has specialized processes, terminology, quality requirements, or output formats that the defaults do not cover.
Build a Skill
A skill is a concise set of instructions that tells Hindsight’s Agent how to complete one well-defined task. It can include the goal, the context to use, the steps to follow, quality checks, and the format of the final output.
In Hindsight, a skill can be selected by name or loaded dynamically when it matches the task at hand. It is not automatically available to Claude, ChatGPT, or another AI tool.
You can still create skills in Claude or ChatGPT, or export a Hindsight skill for use there. Skills are ultimately Markdown instructions, so you can copy the relevant text between tools. When moving a skill out of Hindsight, adapt any Hindsight-specific tool references or company context to the tools and data available in the destination.
Start with one job
Give each skill a narrow, recognizable responsibility. Good examples include:
- Summarize new competitive mentions from this week’s calls
- Prepare a win/loss readout for a quarterly review
- Create an account-specific deal strategy from the latest activity
- Turn buyer feedback into product requests and supporting evidence
Avoid combining unrelated jobs into one skill. A focused skill is easier to select, evaluate, and improve.
Use this structure
Use the following pattern when authoring a skill:
Make skills easy to select
Name skills after the outcome they create, not a vague category. For example, use Prepare Deal Strategy instead of Sales Helper, or Weekly Competitor Brief instead of Competitive Intelligence.
Include a short “When to use it” section with the phrases and scenarios that should trigger it. This helps people select a skill by name and helps AI tools load it dynamically when the request fits.
Give Hindsight the context it needs
Skills are optional, but they give you control over the context and conventions Hindsight should apply. Use them to document information that may not be obvious from an individual request, including:
- When your fiscal quarters start and end
- The currency and reporting conventions your team uses
- The deal fields, segments, and taxonomies that matter for analysis
- The definitions of stages, customer tiers, regions, or other company-specific terms
- The knowledge-base files to use for approved messaging, enablement, or seller guidance
You can also name the specific Hindsight tools the Agent should use. For example, an enablement skill can instruct the Agent to consult a particular knowledge-base file before producing a seller-facing talk track, while an analysis skill can require a particular set of fields for every segmentation.
Test and improve
Run the skill against a few representative requests before sharing it. Check that it:
- Uses the right source material
- Produces specific conclusions rather than generic summaries
- Separates evidence from assumptions
- Follows the requested format every time
If the output is inconsistent, tighten the instructions or add an example of the expected result. Keep the skill’s instructions self-contained rather than relying on undocumented team knowledge.
If a skill depends on Hindsight data, say which tools, fields, taxonomy, or knowledge-base files to use. This makes the output more consistent and gives the Agent the context it needs to follow your team’s process.
Create and customize skills in Hindsight
Go to Skills (/skills) to manage the skills available to your workspace. The page has three libraries:
- Hindsight-created — Skill templates provided by Hindsight.
- Your organization — Skills teammates have created or added for the workspace.
- Yours — Skills you created or added yourself.
Create a skill from scratch
- Select Create skill on the Skills page.
- Add a descriptive name, a short description, category, and the skill prompt.
- Include the goal, relevant Hindsight tools and company context, procedure, quality bar, and output format described above.
- Select Create Skill, then test it in Chat or add it to a workflow.
Add a Hindsight skill to your library
- On Skills, open the Hindsight-created tab.
- Open a template skill and select Add Skill.
- The copied skill appears in Yours, where it is available in your workspace’s skill selectors.
- Open the copy and select Edit to customize its name, description, prompt, and category.
Adding a template creates a copy rather than changing Hindsight’s original. Your customizations affect your workspace’s version only. You can also remix a teammate’s organization skill to make your own editable copy.
If a skill is already used in a workflow, remix it before changing it. This preserves the existing workflow and gives you a safe version to customize.
Create a Workflow
Use a workflow for work that should happen on a recurring schedule: a weekly brief, monthly trend report, regular competitive monitoring, or a periodic update to a team asset.
A workflow is a scheduled prompt. It can contain all the instructions it needs directly, or it can call on a skill for a reusable part of the work.
Set up the prompt
- Go to Workflows in Hindsight and select Create Workflow.
- Give the workflow a clear name that describes its recurring outcome.
- Write the prompt that should run each time.
- Optionally use Browse Skills to add a skill inline, or write the workflow prompt directly.
- Test the prompt and refine the expected output before scheduling it.
The workflow owns its prompt. Adding a skill brings the relevant instructions into that prompt, so a workflow can also stand alone without a linked skill.
Choose a cadence and destination
Select the schedule that matches how quickly the information changes—such as daily, weekly, monthly, or a custom cadence. Then choose where the result should go:
- Slack channel or direct message
- Email recipients
- A new or updated asset in the Hindsight Library
- Zapier, for follow-on actions in other tools
Recurring workflow ideas
- Weekly competitor brief: Review recent calls and surface competitor mentions, objections, and suggested response updates.
- Monthly win/loss report: Identify recurring reasons deals were won or lost, with supporting examples and recommended actions.
- Pipeline risk digest: Summarize open-deal risks, missing next steps, and accounts that need attention.
- Product feedback roundup: Group recent buyer feedback into themes, feature requests, and evidence for the product team.
Use skills outside Hindsight
Hindsight skills run only in Hindsight’s web interface, Responses API, and Slack. To use the same procedure in Claude, ChatGPT, or another AI tool, export or copy the Markdown instructions and add them to that tool’s skill or project setup.
Review the exported instructions before using them elsewhere. Replace Hindsight tool names with the destination tool’s capabilities, and provide any necessary company schema, taxonomy, or source documents that would otherwise have been available in Hindsight.
When to use which
Use a skill when someone needs to run a repeatable task on demand, or when you want an AI tool to recognize and apply specialized instructions. Use a workflow when the task should run without manual effort on a schedule. Combine them when a recurring workflow benefits from a tested, reusable skill—but do not create a skill solely because a workflow needs a prompt.
Best practices
- Keep the task and output specific.
- Give skills descriptive names and clear selection cues.
- Ask for source-backed findings and call out uncertainty.
- Review early workflow runs before relying on automated delivery.
- Revisit prompts and skills as your data, process, or priorities change.
