You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Create an agent skill that knows how to work with OST content #2
templates (reference and/or new cli script commands to generate new OST files
schema definition
validating after edits
identifying missing or misaligned information (eg resolve active opportunities with inactive goals)
Strategic (AI assisted)
Assess entity quality - eg for an opportunity in explore status is the problem statement strong
Workflow and coaching flow for developing and improving the clarity of OST content - challenge the coherence of a ladder from goal to opportunity ("will this really help?"), capture risks and assumptions , prioritisation for discovery and planning
Automate update of knowledge based on other sources (eg decision from meeting notes updates an assumption about a goal, a continuous discovery activity adds insight to a opportunity thesis)
This is probably multiple skills that work together. We already have one that can tactically manage the structure of an OST. We also have the idea of strategic coaching. If we break that down, it would appear to make sense that there are bodies of knowledge that are worth distilling into distinct skills that work together. So for example, we could distil the guidance on opportunity solution trees into one skill. We could have a skill that helps with strategic coaching and that could draw from the various leading thought leaders on strategy. So things like the winning aspiration where we play, how we win. That strategy is about a set of decisions that the layers of a strategy need to be coherent and make sense together that sort of thing that could make for a good strategy coach. However, I also have in mind that there is a an element of tact where a consultant skill rather than coming in blindly and trying to coach people on. Best, practises it might lean on those other skills and apply them by doing so. With awareness that the context of an organisation may not be as ready for all the ideas and it might figure out a way to take an organisation on that journey and pick wedge battles to fight and when in terms of improving their skills around strategy, for example. By doing this, we can create an interesting and diverse range of the skills that work together, but also provide a bit of tension between ideals. One big question of my mind is how much this belongs in this repository versus somewhere else. I have in mind also applications that build on top of these tools and they probably don't belong here either
Following research on #3 (findings in docs/torres-alignment-research.md):
The schema-level descriptive/executable rules proposed in #16 create a natural interface layer for agent skills. Rather than baking qualitative guidance into agent prompts, the schema can carry structured rules per entity type — for example, what makes a valid opportunity (framed in customer voice, grounded in research, in the problem space, not a solution in disguise). An agent can query the schema for these rules when assessing content quality.
This means the schema evolves from pure structural validation into a richer specification that encodes both "is this file structurally valid?" and "is this content well-formed?". The agent skill design and schema design can evolve together, with the schema as the inspectable, shareable source of truth for both.
The tactical tooling skill is substantially complete:
skills/ost-tools/ — a skill file that triggers on ost-tools tasks, covering config discovery, orientation commands, working with spaces, non-obvious issues, and troubleshooting
Agent-friendly commands — spaces and schemas show provide structured, machine-readable output for config inspection and schema exploration
Qualitative content assessment (Add qualitative content assessment to the agent skill #37) — bundled schemas have no rules yet, so agents can't use schemas show to retrieve content quality guidance. That closes the loop on the vision from the second comment here.
Activity
mindsocket commentedon Feb 26, 2026
This is probably multiple skills that work together. We already have one that can tactically manage the structure of an OST. We also have the idea of strategic coaching. If we break that down, it would appear to make sense that there are bodies of knowledge that are worth distilling into distinct skills that work together. So for example, we could distil the guidance on opportunity solution trees into one skill. We could have a skill that helps with strategic coaching and that could draw from the various leading thought leaders on strategy. So things like the winning aspiration where we play, how we win. That strategy is about a set of decisions that the layers of a strategy need to be coherent and make sense together that sort of thing that could make for a good strategy coach. However, I also have in mind that there is a an element of tact where a consultant skill rather than coming in blindly and trying to coach people on. Best, practises it might lean on those other skills and apply them by doing so. With awareness that the context of an organisation may not be as ready for all the ideas and it might figure out a way to take an organisation on that journey and pick wedge battles to fight and when in terms of improving their skills around strategy, for example. By doing this, we can create an interesting and diverse range of the skills that work together, but also provide a bit of tension between ideals. One big question of my mind is how much this belongs in this repository versus somewhere else. I have in mind also applications that build on top of these tools and they probably don't belong here either
mindsocket commentedon Feb 26, 2026
Following research on #3 (findings in
docs/torres-alignment-research.md):The schema-level descriptive/executable rules proposed in #16 create a natural interface layer for agent skills. Rather than baking qualitative guidance into agent prompts, the schema can carry structured rules per entity type — for example, what makes a valid opportunity (framed in customer voice, grounded in research, in the problem space, not a solution in disguise). An agent can query the schema for these rules when assessing content quality.
This means the schema evolves from pure structural validation into a richer specification that encodes both "is this file structurally valid?" and "is this content well-formed?". The agent skill design and schema design can evolve together, with the schema as the inspectable, shareable source of truth for both.
mindsocket commentedon Mar 9, 2026
What was built
The tactical tooling skill is substantially complete:
skills/ost-tools/— a skill file that triggers on ost-tools tasks, covering config discovery, orientation commands, working with spaces, non-obvious issues, and troubleshootingspacesandschemas showprovide structured, machine-readable output for config inspection and schema explorationreferences/schema-authoring.md,references/schema-design.md,references/commands.md~/.local/→$XDG_CONFIG_HOME/, default~/.config/)Tracked separately
schemas showto retrieve content quality guidance. That closes the loop on the vision from the second comment here.