Skills in CC have been a bit frustrating for me. They don't trigger reliably and the emphasis on "it's just markdown" makes it harder to have them reliably call certain tools with the correct arguments.
The idea that agent harnesses should primarily have their functionality dictated by plaintext commands feels like a copout around programming in some actually useful, semi-opinionated functionality (not to mention that it makes capability-discoverability basically impossible). For example, Claude Code has three modes: plan, ask about edits, and auto-accept edits. I always start with a plan and then I end up with multiple tasks. I'd like to auto-accept edits for a step at a time and the only way to do that reliably is to ask CC to do that, but it's not reliable—sometimes it just continues to go into the next step. If this were programmed explicitly into CC rather than relying on agent obedience, we could ditch the nondeterminism and just have a hook on task completion that toggles auto-complete back to "off."
> sometimes it just continues to go into the next step
Use a structured workflow that loops on every task and includes a pause for user confirmation at the end. Enforce it with a hook. I'm not sure if you can toggle auto-accept this way, but I think the end result is what you're asking for.
I use this with great success, sometimes toggling auto-accept on when confidence is high that Claude can complete a step without guidance, and toggling off when confidence is low and you want to slow down and steer, with Claude stopping between the steps. Now that prompt suggestions are a thing, you can just hit enter to continue on the suggested prompt to continue.
> idea that agent harnesses should primarily have their functionality dictated by plaintext commands feels like a copout
I think it's more along the lines of acknowledging the fast-paced changes in the field, and refusing to cast into code something that's likely to rapidly evolve in the near future.
Once things settle down into tested practices, we'll see more "permanent" instrumentation arise.
I think unless you're doing simple tasks, skills are unreliable. For better reliability, I have the agent trigger APIs that handles the complex logic (and its own LLM calls) internally. Has anyone found a solid strategy for making complex 'skills' more dependable?
My only strategy is what used to be called slash-commands but are also skills now, I.e I call them explicitly. I think that actually works quite well and you can allow specific tools and tell it to use specific hooks for security of validation in the frontmatter properties.
So far my experience with skills is that they slow down or confuse agents unless you as the user understand what the skill actually contains and how it works. In general I would rather install a CLI tool and explain to the agent how I want it used vs. trying to get the agent to use a folder of instructions that I don't really understand what's inside.
Most LLM "harnessing" seems very lazy and bolted on. You can build much more robustly by leveraging a more complex application layer where you can manage state, but I guess people struggle building that
Skills feel analogous to behavioral programs. If you give an agent access to a programmable substrate (e.g. bash + CLI tools), you write these Markdown programs which are triggered and read when the agent thinks certain behaviors will be beneficial.
It's a great idea: really neat take on programmability, and can be reloaded while the agent is running without tweaking the harness, etc -- lots of benefits.
`pi` has a great skills implementation too.
I think skills might really shine if you take a minimal approach to the system prompt (like `pi`) -- a lot of the times, if I want to orchestrate the agent in some complex behavior, I want to start fresh, and having it walk through a bunch of skills ... possibly the smaller the system prompt, the more likely the agent is to follow the skills without issue.
the standardization angle is interesting but the real value is in skills that bundle executable scripts alongside the markdown. pure instruction-based skills are fragile because the agent has to interpret intent each time. when the skill includes a concrete shell script or python tool the agent just needs to know when to call it and what args to pass - way more deterministic.
we settled on a pattern where the SKILL.md mostly just describes the interface to a script in the same folder. the agent rarely drifts when it has a concrete tool to invoke vs trying to follow multi-step prose instructions.
> is there a mechanism to pin a version, or is it always HEAD? Skills that evolve can silently break downstream workflows.
don't forget these skills are just text that goes into the llm for it to read, interpret, and then produce text that then gets executed in bash. The more intricate and specific the skill definition the more likely the model is to miss something or not follow it exactly.
I actually think SKILLS.md is such a janky way of doing this sort of thing, let alone the fact that's reliant on the oh-so-brittle Python ecosystem. Also way too much context/tokens being eaten up by something that could be piece-wise programmatically injected in the token stream.
Except they're still not accepting any feedback around AGENTS.md as a standard. You need to explicitly symlink CLAUDE.md to AGENTS.md in a workspace in order to Claude to work like every other agent when it comes to loading context.
The idea that agent harnesses should primarily have their functionality dictated by plaintext commands feels like a copout around programming in some actually useful, semi-opinionated functionality (not to mention that it makes capability-discoverability basically impossible). For example, Claude Code has three modes: plan, ask about edits, and auto-accept edits. I always start with a plan and then I end up with multiple tasks. I'd like to auto-accept edits for a step at a time and the only way to do that reliably is to ask CC to do that, but it's not reliable—sometimes it just continues to go into the next step. If this were programmed explicitly into CC rather than relying on agent obedience, we could ditch the nondeterminism and just have a hook on task completion that toggles auto-complete back to "off."
Use a structured workflow that loops on every task and includes a pause for user confirmation at the end. Enforce it with a hook. I'm not sure if you can toggle auto-accept this way, but I think the end result is what you're asking for.
I use this with great success, sometimes toggling auto-accept on when confidence is high that Claude can complete a step without guidance, and toggling off when confidence is low and you want to slow down and steer, with Claude stopping between the steps. Now that prompt suggestions are a thing, you can just hit enter to continue on the suggested prompt to continue.
I think it's more along the lines of acknowledging the fast-paced changes in the field, and refusing to cast into code something that's likely to rapidly evolve in the near future.
Once things settle down into tested practices, we'll see more "permanent" instrumentation arise.
It's a great idea: really neat take on programmability, and can be reloaded while the agent is running without tweaking the harness, etc -- lots of benefits.
`pi` has a great skills implementation too.
I think skills might really shine if you take a minimal approach to the system prompt (like `pi`) -- a lot of the times, if I want to orchestrate the agent in some complex behavior, I want to start fresh, and having it walk through a bunch of skills ... possibly the smaller the system prompt, the more likely the agent is to follow the skills without issue.
we settled on a pattern where the SKILL.md mostly just describes the interface to a script in the same folder. the agent rarely drifts when it has a concrete tool to invoke vs trying to follow multi-step prose instructions.
don't forget these skills are just text that goes into the llm for it to read, interpret, and then produce text that then gets executed in bash. The more intricate and specific the skill definition the more likely the model is to miss something or not follow it exactly.
For example, on Proposal: AgentFile — Declarative Agent Composition from Skills + Filesystem-Native Skill Delivery
https://github.com/agentskills/agentskills/discussions/179
Imo a bad idea, but alas.
Isn’t python just an option ?
1. Using the skills frontmatter to implement a more complex YAML structure, so e.g.
2. Using a skills lock file ;-)